394 replies
19h29m
Well, that statement lays out a damning timeline:
- OpenAI approached Scarlett last fall, and she refused.
- Two days before the GPT-4o launch, they contacted her agent and asked that she reconsider. (Two days! This means they already had everything they needed to ship the product with Scarlett’s cloned voice.)
- Not receiving a response, OpenAI demos the product anyway, with Sam tweeting “her” in reference to Scarlett’s film.
- When Scarlett’s counsel asked for an explanation of how the “Sky” voice was created, OpenAI yanked the voice from their product line.
Perhaps Sam’s next tweet should read “red-handed”.
This statement from scarlet really changed my perspective. I use and loved the Sky voice and I did feel it sounded a little like her, but moreover it was the best of their voice offerings. I was mad when they removed it. But now I’m mad it was ever there to begin with. This timeline makes it clear that this wasn’t a coincidence and maybe not even a hiring of an impressionist (which is where things get a little more wishy washy for me).
The thing about the situation is that Altman is willing to lie and steal a celebrity's voice for use in ChatGPT. What he did, the timeline, everything - is sleazy if, in fact, that's the story.
The really concerning part here is that Altman is, and wants to be, a large part of AI regulation [0]. Quite the public contradiction.
[0] https://www.businessinsider.com/sam-altman-openai-artificial...
Altman has proven time and again that he is little more than a huckster wrt technology, and in business he is a stone cold shark.
Conman plain and simple.
You'd think that Worldcoin would be enough proof of what he is but I guess people missed that memo.
Much as I dislike crypto, that's more of "having no sense of other people's privacy" (and hubris) than general scamminess.
It's a Musk-error not an SBF-error. (Of course, I do realise many will say all three are the same, but I think it's worth separating the types of mistakes everyone makes, because everyone makes mistakes, and only two of these three also did useful things).
It's not just about privacy either.
Worldcoin is centrally controlled making it a classic "scam coin". Decentralization is the _only_ unique thing about cryptocurrencies, when you abandon decentralization all that's left is general scamminess.
(Yes, there's nuance to decentralization too but that's not what's going on with Worldcoin.)
True decentralisation is part of the problem with cryptocurrencies and why they can't work the way the advocates want them to.
Decentralisation allows trust-less assurance that money is sent, it's just that's not useful because the goods or services for which the money is transferred still need either trust or a centralised system that can undo the transaction because fraud happened.
That's where smart contracts come in, which I also think are a terrible idea, but do at least deserve a "you tried!" badge, because they're as dumb as saying "I will write bug-free code" rather than as dumb as "let's build a Dyson swarm to mine exactly the same amount of cryptocurrency as we would have if we did nothing".
That is indeed something it does.
But it also gives you the assurance that a single entity can't print unlimited money out of thin air, which is the case with a centrally controlled currency like Worldcoin.
They can just shrug their shoulders and claim that all that money is for the poor and gullible Africans that had their eyeballs scanned.
Sure, but the inability to do that when needed is also a bad thing.
Also, single world currencies are (currently) a bad thing, because when your bit of the world needs to devalue its currency is generally different to when mine needs to do that.
But this is why economics is its own specialty and not something that software nerds should jump into like our example with numbers counts for much :D
When and why would BTC or ETH need to print unlimited money and devalue themselves?
Wrong framing, currencies don't have agency. You should be asking when would you need your currency to be devalued, regardless of what it's called or made from.
And the answer to that is all the reasons governments do just that, except for the times where the government is being particularly stupid and doing hyperinflation.
Not a very convincing answer at all.
What would a convincing answer look like?
Sufficiently advanced incompetence is indistinguishable from malice.
It's not particularly advanced, it's the same thing that means the supermajority of websites have opted for "click here to consent to our 1200 partners processing everything you do on our website" rather than "why do we need 1200 partners anyway?"
It's still bad, don't get be wrong, it's just something I can distinguish.
If it fools billions of people and does significant damage to the lives of people, then it's plenty advanced to me, even if it happens through a more simple or savant-like process than something that looks obviously deliberate.
I don't think the cookies thing is a good example. That's passive incompetence, to avoid the work of changing their business models. Altman actively does more work to erode people's rights.
Can you? Plausible deniability is one of the first things in any malicious actor's playbook. "I meant well…" If there's no way to know, then you can only assess the pattern of behavior.
But realistically, nobody sapient accidentally spends multiple years building elaborate systems for laundering other people's IP, privacy, and likeness, and accidentally continues when they are made aware of the harms and explicitly asked multiple times to stop…
It’s both.
https://news.ycombinator.com/item?id=40427454
Because of course he's got a crypto grift going. Shocking.
I think people tend to assume our own values and experiences have some degree of being universal.
So scammers see other scammers, and they just think there's nothing wrong with it.
While normal people who act in good faith see scammers, and instinctively think that there must be a good reason for it, even (or especially!) if it looks sketchy.
I think this happens a lot. Not just with Altman, though that is a prominent currently ongoing example.
Protecting yourself from dark triad type personalities means you need to be able to understand a worldview and system of values and axioms that is completely different from yours, which is… difficult. …There's always that impulse to assume good faith and rationalize the behavior based on your own values.
Not going to lie, he had me. He appeared very genuine and fair in almost all media he appeared like podcasts but many of his actions are just so hard to justify.
He has a certain charm and seeming sincerity when he talks. But the more I see of him, the more disturbing I find him -- he combines the Mark Zuckerberg stare with the Elizabeth Holmes vocal fry.
so all psychopaths do, aren't they?
CEO's have been studied to have a disproportionately higher rate of psychopathy. So there's a little correlation. You don't get to the top of a company in this kind of society without having some inherent charm (assuming you aren't simply inheriting billions from a previous generation).
Do you have a link to a video of Altman's voice shifting from controlled deep to nasal? The videos of Elizabeth Holmes not being able to keep up with the faked deep tone are textbook-worthy...
I have exactly the same feeling as I think you do. When you reach the levels of success he has, there will always be people screaming that you are incompetent, evil and every other negative adjective under the sun. But he genuinely seemed to care about doing the right thing. But this is just so lacking of basic morals that I have to conclude that I was wrong, at least to an extent.
I feel that this is a classic tale of success getting to you. It almost feels like it's impossible to be successful at this level and remain true. At least, I hadn't seen it yet.
Or "success" itself acts as a filter selecting for those who are ruthless enough to do amoral and immoral things.
He just, to his credit, understands his public persona has to be the non-douchy-tech-bro and the media will eat it up. Much like a politician. He doesn't want to be like Elon or Travis K in public (though he probably agrees with them more than his public persona would imply).
This is why it's a mistake to go by "vibes" of a person when they're speaking to an audience. Pay attention to what they do, not what they say.
I'm glad more people are thinking this. It's amazing that he got his way back into OpenAI somehow. I said as much that he shouldn't go back to OpenAI and got downvotes universally both here and on reddit.
Aspiring technofeudalist.
almost every tech ceo is like that. Could list many examples. It's an effect of capitalism.
if this account is true, Sam Altman is a deeply unethical human being. Given that he doesn't bring any technical know how to building of AGI, I just don't see the reason to have such a person in charge here. The new board should act.
He has “The Vision”… It’s the modern entrepreneurship trope that lowly engineers won’t achieve anything if they weren’t rallied by a demi-god who has “The Vision” and makes it all happen.
Probably not wrong. Lots and lots of examples of that being true.
There is something to it. Someone has to identify the intersection between what the engineering can do and what the market actually wants, then articulate that to a broad enough audience. Engineers constantly undervalue this very fuzzy and very human centric part of the work.
I don't think the issue is that Vision doesn't matter. I think the issue is Sam doesn't have it. Like Gates and Jobs had clear, well defined visions for how the PC was going to change the world, then rallied engineering talent around them and turned those into reality, that's how their billions and those lasting empires were born. Maybe someone like Elon Musk is a contemporary example. Just don't see anything like that from SamA, we see him in the media, talking a lot about AI, rubbing shoulders with power brokers, being cutthroat, but where's the vision of a better future? And if he comes up with one does he really understand the engineering well enough to ground it in reality?
I don’t know enough about him or his vision. It doesn’t seem he’s as clear as say Jobs in the past. But I do look at all the amazing things openai has done in a short period of time, and that the employees overwhelmingly backed him with the whole board chaos issue. He also has fundraised a lot money for the company. It appears he’s doing more right than wrong, and openai pulled everyone else’s pants down.
I roll my eyes when somebody says that they’re “the idea person” or that they have “the vision”.
I’d wager that most senior+ engineers or product people also have equally compelling “the vision”s.
The difference is that they need to do actual work all day so they don’t get to sit around pontificating.
Ideas are a dime a dozen. The value of "idea men" isn't their ideas, it's their ability to rally people around them. It's the exact same skill that con men use for nefarious purposes.
It shouldn’t be forgotten that his sister has publicly accused him and his brother of sexually abusing her as a child.
I didn't know about that, strange:
https://www.lesswrong.com/posts/QDczBduZorG4dxZiW/sam-altman...
"Some commenters on Hacker News claim that a post regarding Annie's claims that Sam sexually assaulted her at age 4 has been being repeatedly removed."
Whelp. Let us see if this one sticks.
He must be bringing something to the table as they tried to get rid of him and failed spectacularly. Business is not only about technical know how.
Microsoft. They are protecting their investment.
I mean, there's already been some yellow flags with Altman already. He founded Worldcoin, whose plan is to airdrop free money in exchange for retinal scans. And the board of OpenAI fired him for (if I've got this right) lying to the board about conversations he'd had with individual board members.
WorldCoin is how I first heard of him, and it's what made me think he was a bad actor. I think of it as a red flag, not yellow.
He rubs elbows with very powerful people including CEOs, heads of state and sheiks. They probably want 'one of them' in charge of the company that has the best chances of getting close to AGI. So it's not his technical chops and not even 'vision' in the Jobs sense that keeps him there.
Are they really the ones with the best chance now though?
They're basically owned by Microsoft, they're bleeding tech/ethnical talent and credibility, and most importantly Microsoft Research itself is no slouch (especially post-Deepmind poaching) - things like Phi are breaking ground on planets that openai hasn't even touched.
At this point I'm thinking they're destined to become nothing but a premium marketing brand for Microsoft's technology.
This isn’t even close to the most unethical thing he has done. This is peanuts compared to the Worldcoin scam.
https://news.ycombinator.com/item?id=40427454
I thought we had already established this when the previous board tried to oust him for failing to stick to OpenAI’s charter. This is just further confirmation.
You mean like the last board tried? Besides the board was picked to be on Altman’s side. The independent members were forced out.
because so many people ran cover for him, from paul graham to whos-who of silicon valley.
Most likely it was an unforced error, as there’ve been a lot of chaos with cofounders and the board revolt, easy to loose track of something really minor.
Like some intern’s idea to train the voice on their favorite movie.
And then they’ve decided that this is acceptable risk/reward and not a big liability, so worth it.
This could be a well-planned opening move of a regulation gambit. But unlikely.
This is an unforced error, but it isn’t minor. It’s quite large and public.
The general public doesn’t understand the details and nuances of training an LLM, the various data sources required, and how to get them.
But the public does understand stealing someone’s voice. If you want to keep the public on your side, it’s best to not train a voice with a celebrity who hasn’t agreed to it.
I had a conversation with someone responsible for introducing LLMs into the process that involves personal information. That person rejected my concern over one person's data appearing in the report on another person. He told me that it will be possible to train AI to avoid that. The rest of the conversation convinced me that AI is seen as magic that can do anything. It seems to me that we are seeing a split between those who don't understand it and fear it and those who don't understand it, but want to align themselves with it. Those latter are those I fear the most.
The "AI is magic and we should simply believe" is even being actively promoted because all these VC hucksters need it.
Any criticism of AI is being met with "but if we all just hype AI harder, it will get so good that your criticisms won't matter" or flat out denied. You've got tech that's deeply flawed with no obvious way to get unflawed, and the current AI 'leaders' run companies with no clear way to turn a profit other than being relentlessly hyped on proposed future growth.
It's becoming an extremely apparent bubble.
On the plus side, lots of cheap nVida cards heading for eBay once it bursts.
I don't think this makes any sense, at all, quite honestly. Why would an "intern" be training one of ChatGPT's voices for a major release?
If in fact, that was the case, then OpenAI is not aligned with the statement they just put out about having utmost focus on rigor and careful considerations, in particular this line: "We know we can't imagine every possible future scenario. So we need to have a very tight feedback loop, rigorous testing, careful consideration at every step, world-class security, and harmony of safety and capabilities." [0]
[0] https://x.com/gdb/status/1791869138132218351
It makes a lot more sense that he was caught red-handed, likely hiring a similar voice actress and not realizing how strong identity protections are for celebs.
Ah, the famous rogue engineer.
The thing is, even if it were the case, this intern would have been supervised by someone, who themselves would have been managed by someone, all the way to the top. The moment Altman makes a demo using it, he owns the problem. Such a public fuckup is embarrassing.
You mean, they were reckless and tried to wing it? Yes, that’s exactly what’s wrong with them.
LOL. ROFL, even. This was a gambit all right. They just expected her to cave and not ask questions. Altman has a common thing with Musk: he does not play 3D chess.
Yes, because we all know the high profile launch for a major new product is entirely run by the interns. Stop being an apologist.
At first I thought there may be a /s coming...
He lies and steals much more than that. He’s the scammer behind Worldcoin.
https://www.technologyreview.com/2022/04/06/1048981/worldcoi...
https://www.buzzfeednews.com/article/richardnieva/worldcoin-...
That’s as much of a contradiction as a thief wanting to be a large part of lock regulation. What better way to ensure your sleazy plans benefit you, and preferably only you but not the competition, than being an active participant in the inevitable regulation while it’s being written?
Based on what I see in the videos from The Lockpocking Lawyer, that would be a massive improvement.
Now, the NSA and crypto standards, that would have worked as a metaphor for your point.
(I don't think it's correct, but that's an independent claim, and I am not only willing to discover that I'm wrong about their sincerity, I think everyone writing that legislation should actively assume the worst while they do so).
The Lockpicking Lawyer is not a thief, so I don’t get your desire to incorrectly nitpick. Especially when you clearly understood the point.
You noticed your confusion but still went on the aggressive, huh. Ah well.
"A is demonstrating a proof of B" does not require "A is a clause in B".
A being TLPL, B being that the entire lock industry is bad, so bad that anyone with experience would be a massive improvement, for example a thief.
I’m not confused and my reply was not aggressive. I don’t think it will be a good use of time to continue this conversation because discussions should get more substantive as they go on and this was an irrelevant tangent to which I have no desire to get sucked in to.
Other people have commented to further explain the point in other words. I recommend you read those, perhaps it’ll make you understand.
https://news.ycombinator.com/item?id=40428005
https://news.ycombinator.com/item?id=40428280
A thief is not a lock picker and they don't have the same incentive. A thief in a position to dictate lock regulation would try to have a legal backdoor on every lock in the world. One that only he has the master key for. Something something NSA & cryptography :)
Indeed, but locks are so bad (as demonstrated by LPL) that even a thief would make them better.
Indeed, as I said :)
If you've watched his videos then surely you should know that lockpicking isn't even on the radar for thieves as there are much easier and faster methods such as breaking the door or breaking a window.
Surely that just makes the thief/locks metaphor even worse? Or have I missed something?
Some people might see some parallel with SBF and see how Altman would try to regulate competition without impeding OpenAI progress
I always mix up those two in my head and have to think which one is which
One is in jail, when it should be two are in jail
I dont like sam, but he moves way smarter than ppl like sbf or Elizabeth holmes. He actual has a product close to the reported specs, albeit still far away from the ultimate goal of AGI
i dont see why he should be in jail
If his sister's words about sexually abusing her are true, he should be in jail.
no, in that case he should have been in the Juvenile incarceration system, unless the argument is that he should have been charged as an adult, or that Juvenile abusers should always be charged and sentenced as adults, or that Juvenile sex offenders who were not charged as Juveniles should be charged as adults.
Which one?
on edit: this being based on American legal system, you may come from a legal system with different rules.
Should be in jail for Worldcoin which has pilfered people of their biological identity. I guess you could literally delete Worldcoin and in theory make people whole, but that company treats humans like vegetables that have no rights.
Maybe they were the rogue AGI escapes we found along the way
sama gets to farm out much of the lobbying to Microsoft’s already very powerful team, which spends a mere $10m but that money gets magnified by MS’s gov and DoD contracts. That’s a huge safety net for him, he gets to steal and lie (as demonstrated w/ Scarlett) and yet the MS lobbying machine will continue unphased.
https://www.opensecrets.org/federal-lobbying/clients/summary...
the whole technology is based on fucking over artists, who didn't expect this exact thing?
It's not just the artists, anything you do in the digital realm and anything that can be digitised is fair game. In the UK NHS GP practices refuse to register you to see a doctor even when it's urgent and tell you to use a third-party app to book an appointment. You have use your phone to take photos of the affected area and provide a personal info. I fully expect that data to be fed into some AI and sold without me knowing and without a process for removal of data should the company go bust. It is preying on the vulnerable when they need help.
Important to note the "The NHS" is not a single entity and the GP practice is likely a private entity owned in partnership by the doctors. There are a number of reasons why individual practices can refuse to register.
Take your point about LLMs though.
I went to see my GP and the lady at the reception told me they no longer book visits at the reception and I had to use the app. Here's the privacy policy https://support.patientaccess.com/privacy-policy They reserve the right to pass your data to third party contractors and to use it for marketing purposes. There is the obligatory clause on regarding the right to be forgotten, but the AI companies claim it is impossible to implement.
I didn't read that as reserving the right - looks like a standard dpia that is opt-in and limited.
However, GP practices are essentially privatised - so you do have the right to register at another practice.
Last time I booked a blood test it was via the official NHS app , not a third party.
https://www.patientaccess.com/
App? What's an app?
It's a thing you put on your phone
I don't have a phone
Well, we can't register you
You don't accept people who don't have phones? Could I have that in writing please, ..., oh, your signature on that please ...
Altman wants to be a part of AI regulation in the same way Bankman Fried wanted to be a part of cryptocurrency regulation.
Whats really interesting about our timeline is when you look at the history of market capture in Big Oil, Telco, Pharma, Real Estate, Banks, Tobacco etc all the lobbying, bribing, competition killing used to be done behind the scenes within elite circles.
The public hardly heard from or saw the mgmt of these firm in media until shit hit the fan.
Today it feels like managment is in the media every 3 hours trying to capture attention of prospective customers, investors, employees etc or they loose out to whoever is out there capturing more attention.
So false and condradictory signalling is easy to see. Hopefully out of all this chaos we get a better class of leaders not a better class of panderers.
So great to have twitter so the narcissistic psychopaths can't resist revealing themselves for clout.
I mean, to whatever extent it matters. All these outrageously rich morons still have tons of economic and social clout. They still have pages upon pages of fans foaming at the mouth for the opportunity to harass people asking basic questions. They still carry undue influence in our society and in our industry no matter how many times they are "outed."
What does being outed even mean anymore? It's just free advertising from all the outlets that feel they can derive revenue off your name being in their headlines. Nothing happens to them. SBF and Holmes being the notable exceptions, but that's because they stole from rich people.
(AI)tman tries to be, (Bank)man fried to be, who is letting Kojima name all these villians?
I always had trouble telling apart those two Sams. Turns out they're the same person.
What is so special about her voice? They could’ve found a college student with a sweet voice and offered to pay her tuition in exchange for using her voice, no? Or a voice actor?
Why be cartoonishly stupid and cartoonishly arsehole and steal a celebrity’s voice? Did he think Scarlett won’t find out? Or object?
I don’t understand these rich people. Is it their hobby to be a dick to as many people as they can, for no reason other than their amusement? Just plain weirdos
Scarlett voiced Samantha, an AI in the movie "Her"
Considering the movie's 11 years old, it's surprisingly on-point with depictions of AI/human interactions, relations, and societal acceptance. It does get a bit speculative and imaginative at the end though...
But I imagine that movie did/does spark the imagination of many people, and I guess Sam just couldn't let it go.
It's not just that. Originally the AI voice in Her was played by someone else, but Spike Jonze felt strongly that the movie wasn't working and recast the part to Johansson. The movie immediately worked much better and became a sleeper hit. Johansson just has a much better fitting voice and higher skill in voice acting for this kind of role, to the extent that it maybe was a make/break choice for the movie. It isn't a surprise that after having created the exact tech from the movie, OpenAI wanted it to have the same success that Jonze had with his character.
It's funny that just seven days ago I was speculating that they deliberately picked someone whose voice is very close to Scarlett's and was told right here on HN, by someone who works in AI, that the Sky voice doesn't sound anything like Scarlett and it is just a generic female voice:
https://news.ycombinator.com/item?id=40343950#40345807
Apparently .... not.
They seem to love "testing" how much they can bully someone.
I remember a few experiences where someone responded by being an even bigger dick, and they disappeared fast.
Correcting, the thing about this whole situation with OpenAI is they are willing to steal everything for use in ChatGPT. They trained their model with copyrighted data and for some reason they won't delete the millions of protected data they used to train the AI model.
Using other people data for training without their permission is the "original sin" of LLMs[1]. That will, at best, be a shadow over the entire field for an extremely long time.
[1] Just to head off people saying that such a use is not a copyright violation -- I'm not saying it is. I'm just saying that it's extremely sketchy and, in my view, ethically unsupportable.
Altman is a known conman. Surely you are aware of Yishan Wong describing how Sam Altman and the Reddit founders conned Conde Nast https://reddit.com/r/AskReddit/comments/3cs78i/whats_the_bes...
Wow, Altman in the replies there:
Altman doesn’t want to be part of regulation. sama wants to be the next tk. he wants to be above regulation, and he wants to spend Microsoft’s money getting there.
E.g. flying Congress to Lake Cuomo for an off-the-record “discussion” https://freebeacon.com/politics/how-the-aspen-institute-help...
This whole exchange from 1:04:53 to 1:10:22 takes a whole different meaning....
https://youtu.be/P_ACcQxJIsg?t=3891
"Not consistently candid", the last board said.
Like many people who try to oppose psychopaths though, they don't seem to be around much anymore.
But it's clearly not her voice right? The version that's been on the app for a year just isn't. Like, it clearly intending to be slightly reminiscent of her, but it's also very clearly not. Are we seriously saying we can't make voices that are similar to celebrities, when not using their actual voice?
They clearly thought it was close enough that they asked for permission, twice. And got two no’s. Going forward with it at that point was super fucked up.
It’s very bad to not ask permission when you should. It’s far worse to ask for permission and then ignore the response.
Totally ethically bankrupt.
You seem to be misunderstanding the situation here. They wanted ScarJo to voice their voice assistant, and she refused twice. They also independently created a voice assistant which sounds very similar to her. That doesn't mean they thought they had to ask permission for the similar voice assistant.
You seem to be misunderstanding the legalities at work here: reaching out to her multiple times beforehand, along with tweets intended to underline the similarity to her work on Her, demonstrates intention. If they didn’t think they needed permission, why ask for permission multiple times and then yank it when she noticed?
Answer: because they knew they needed permission, after working so hard to associate with Her, and they hoped that in traditional tech fashion that if they moved fast and broke things enough, everyone would have to reshape around OAs wants, rather than around the preexisting rights of the humans involved.
You could also ask: If Scarlett has a legal case already, why does she want legislation passed?
Before Roe vs Wade was overturned you might have asked if abortion is legal why do abortion rights advocates want legislation passed?
The answer is without legislation you are far more subject to whether a judge feels like changing the law.
Because a legal case under the current justice system and legislative framework would probably take hundreds of thousands to millions of dollars to bring a case that requires discovery and a trial to accomplish.
Maybe (maybe!) it’s worth it for someone like Johansson to take on the cost of that to vindicate her rights—but it’s certainly not the case for most people.
If your rights can only be defended from massive corporations by bringing lawsuits that cost hundreds of thousands to millions of dollars, then only the wealthy will have those rights.
So maybe she wants new legislative frameworks around these kind of issues to allow people to realistically enforce these rights that nominally exist.
For an example of updating a legislative framework to allow more easily vindicating existing rights, look up “anti-SLAPP legislation”, which many states have passed to make it easier for a defendant of a meritless lawsuit seeking to chill speech to have the lawsuit dismissed. Anti-SLAPP legislation does almost nothing to change the actual rights that a defendant has to speak, but it makes it much more practical for a defendant to actually excercise those rights.
So, the assumption that a call for updated legislation implies that no legal protection currently exists is just a bad assumption that does not apply in this situation.
To prevent it from happening again, with more legal authority than a legal precedent.
She has a personal net worth of >$100m. She’s also married to a successful actor in his own right.
Her voice alone didn’t get her there — she did. That’s why celebrities are so protective about how their likeness is used: their personal brand is their asset.
There’s established legal precedent on exactly this—even in the case they didn’t train on her likeness, if it can reasonably be suspected by an unknowing observer that she personally has lent her voice to this, she has a strong case. Even OpenAI knew this, or they would not have asked in the first place.
Many things that are legal are of questionable ethics. Asking permission could easily just be an effort for them to get better samples of her voice. Pulling the voice after debuting it is 100% a PR response. If there's a law that was broken, pulling the voice doesn't unbreak it.
They are trying to wriggle out of providing insight into how that voice was derived at all (like Google with the 100% of damages check). It would really suck for OpenAI if, for example, Altman had at some point emailed his team to ensure the soundalike was “as indistinguishable from Scarlet’s performance in HER as possible.“
Public figures own their likeness and control its use. Not to mention that in this case OA is playing chicken with studios as well. Not a great time to do so, given their stated hopes of supplanting 99% of existing Hollywood creatives.
One very easy explanation is that they trained Sky using another voice (this is the claim and no reason to doubt it is true) wanting to replicate the stye of the voice in "Her", but would have preferred to use SJ's real voice for the PR impact that could have.
Yanking it could also easily be a pre-emptive response to avoid further PR drama.
You will obvious decide you don't believe those explanations, but to many of us they're quite plausible, in fact I'd even suggest likely.
(And none of this precludes Sam Altman and OpenAI being dodgy anyway)
I actually believe that’s quite plausible. The trouble is, by requesting permission in the first place, they demonstrated intent, which is legally significant. I think a lot of your confusion is attempting to employ pure logic to a legal issue. They are not the same thing, and the latter relies heavily on existing precedent — of which you may, it seems, be unaware.
So, what would they have done if she accepted? Claimed that the existing training of the Sky voice was voiced by her?
That claim could very well be true. The letter requested information on how the voice was trained - OpenAI may not want that can of worms opened lest other celebrities start paying closer attention to the other voices.
Voice cloning could be as simple as a few seconds of audio in the context window since GPT-4o is a speech to speech transformer. They wouldn't need to claim anything, just switch samples. They haven't launched the new voice mode yet, just demos.
That would be suspicious too
Maybe they have second trained on her voice.
And promoted it using a tweet naming the movie that Johansson performed in, for the role that prompted them to ask her in the first place.
You have to be almost deliberately naive to not see that the were attempting to use her vocal likeness in this situation. There’s a reason they immediately walked it back after the situation was revealed.
Neither a judge, nor a jury, would be so willingly naive.
This is a genuine question. If it turns out they trained Sky on someone else's voice to similarly replicate the style of the voice in "Her", would you be ok with that? If it was proven that the voice was just similar, to SJ's would that be ok?
My view is, of course it is ok. SJ doesn't own the right to a particular style of voice.
And... No. That is what OpenAI will assert, and good discovery by Scar Jo reps may prove or disprove.
Yes, totally ethically bankrupt. But what bewilders me is that they yanked it as soon as they heard from their lawyers. I would have thought that if they made the decision to go ahead despite getting two "no"s, that they at least had a legal position they thought was defensible and worth defending.
But it kind of looks like they released it knowing they couldn't defend it in court which must seem pretty bonkers to investors.
They likely have a legal position which is defensible.
They're much more worried that they don't have a PR position which is defensible.
What's the point of winning the (legal) battle if you lose the war (of public opinion)?
Given the rest of their product is built on apathy to copyright, they're actively being sued by creators, and the general public is sympathetic to GenAI taking human jobs...
... this isn't a great moment for OpenAI to initiate a long legal battle, against a female movie actress / celebrity, in which they're arguing how her likeness isn't actually controlled by her.
Talk about optics!
(And I'd expect they quietly care much more about their continued ability to push creative output through their copyright launderer, than get into a battle over likeness)
How is the PR position not defensible? One of the worst things you can generally do is admit fault, particularly if you have a complete defense.
Buckle in, go to court, and double-down on the fact that the public's opinion of actors is pretty damn fickle at the best of times - particularly if what you released was in fact based on someone you signed a valid contract with who just sounds similar.
Of course, this is all dependent on actually having a complete defense of course - you absolutely would not want to find Scarlett Johannsen voice samples in file folders associated with the Sky model if it went to court.
In what world does a majority of the public cheer for OpenAI "stealing"* an actress's voice?
People who hate Hollywood? Most of that crowd hates tech even more.
* Because it would take the first news cycle to be branded as that
It is wild to me that on HackerNews of all places, you'd think people don't love an underdog story.
Which is what this would be in the not-stupid version of events: they hired a voice actress for the rights to create the voice, she was paid, and then is basically told by the courts "actually you're unhireable because you sound too much like an already rich and famous person".
The issue of course is that OpenAIs reactions so far don't seem to indicate that they're actually confident they can prove this or that this is the case. Coz if this is actually the case, they're going about handling this in the dumbest possible way.
It’s wild to me that there are people who think that OpenAI are the underdog. A 80Bn Microsoft vassal, what a plucky upstart.
You realise that there are multiple employees including the CEO publicly drawing direct comparisons to the movie Her after having tried and failed twice to hire the actress who starred in the movie? There is no non idiotic reading of this.
You're reading my statements as defending OpenAI. Put on your "I'm the PR department hat" and figure out what you'd do if you were OpenAI given various permutations of the possible facts here.
That's what I'm discussing.
Edit: which is to say, I think Sam Altman may have been a god damn idiot about this, but it's also wild anyone thought that ScarJo or anyone in Hollywood would agree - AI is currently the hot button issue there and you'd find yourself the much more local target of their ire.
Then why bother mentioning an "underdog story" at all?
Who is the underdog in this situation? In your comment it seems like you're framing OpenAI as the underdog (or perceived underdog) which is just bonkers.
Hacker News isn't a hivemind and there are those of us who work in GenAI who are firmly on the side of the creatives and gasp even rights holders.
an obnoxious sleazy millionaire backed by microsoft is by no means “an underdog”
There are quite a few issues here: First, this is assuming they actually hired a voice-alike person, which is not confirmed. Second, they are not an underdog (the voice actress might be, but she's most likely pretty unaffected by this drama). Finally, they were clearly aiming to impersonate ScarJo (as confirmed by them asking for permission and samas tweet), so this is quite a different issue than "accidentally" hiring someone that "just happens to" sound like ScarJo.
Doesn't sound like they have that either.
Copilot still tells me I've commit a content policy violation of I ask it to generate an image "in Tim Burton's style". Tim Burton has been openly critical of generative AI.
That actually seems like there may be a few people involved and one of them is a cowboy PM who said fuck it, ship it to make the demo. And then damage control came in later. Possibly the PM didn't even know about the asks for permission?
Given the timeline it sounds like the PM was told "just go ahead with it, I'll get the permission".
Ken Segall has a similar Steve Jobs story, he emails Jobs that the Apple legal team have just thrown a spanner in the works days before Ken's agency is set to launch Apple's big ad campaign and what should he do?
Jobs responds minutes later... "Fuck the lawyers."
The whole company behaves like rogue cowboys.
If a PM there didn’t say “fuck it ship it even without her permission” they’d probably be replaced with someone who would.
I expect the cost of any potential legal action/settlement was happily accepted in order to put on an impressive announcement.
It looks really unprofessional at minimum if not a bit arrogant, which is actually more concerning as it hints at a deeper disrespect for artists and celebrities.
Effective altruism would posit that it is worth one voice theft to help speed the rate of life saving ai technology in the hands of everyone.
Effective Altruists are just shitty utilitarians that never take into account all the myriad ways that unmoderated utilitarianism has horrific failure modes.
Their hubris will walk them right into federal prison for fraud if they’re not careful.
If Effective Altruists want to speed the adoption of AI with the general public, they’d do well to avoid talking about it, lest the general public make a connection between EA and AI
I will say, when EA are talking about where they want to donate their money with the most efficacy, I have no problem with it. When they start talking about the utility of committing crimes or other moral wrongs because the ends justify the means, I tend to start assuming they’re bad at morality and ethics.
This is like attributing the crimes of a few fundamentalists to an entire religion.
I don’t think so. I’ve narrowed my comments specifically to Effective Altruists who are making utilitarian trade-offs to justify known moral wrongs.
Frankly, if you’re going to make an “ends justify the means” moral argument, you need to do a lot of work to address how those arguments have gone horrifically wrong in the past, and why the moral framework you’re using isn’t susceptible to those issues. I haven’t seen much of that from Effective Altruists.
I was responding to someone who was specifically saying an EA might argue why it’s acceptable to commit a moral wrong, because the ends justify it.
So, again, if someone is using EA to decide how to direct their charitable donations, volunteer their time, or otherwise decide between mora goods, I have no problem with it. That specifically wasn’t context I was responding to.
Did you?
Sure, I should’ve said I tried to or I intended to:
You can see another comment here, where I acknowledge I communicate badly, since I’ve had to clarify multiple times what I was intending: https://news.ycombinator.com/item?id=40424566
This is the paragraph that was intended to narrow what I was talking about:
That said, I definitely should’ve said “those Effective Altruists” in the first paragraph to more clearly communicate my intent.
Effective Altruists are the fundamentalists though. So no, it's not.
Extremely reasonable position, and I'm glad that every time some idiot brings it up in the EA forum comments section they get overwhelmingly downvoted, because most EAs aren't idiots in that particular way.
I have no idea what the rest of your comment is talking about; EAs that have opinions about AI largely think that we should be slowing it down rather than speeding it up.
In some sense I see a direct line between the EA argument being presented here, and the SBF consequentialist argument where he talks about being willing to flip a coin if it had a 50% chance to destroy the world and a 50% chance to make the world more than twice as good.
I did try to cabin my arguments to Effective Altrusts that are making ends justify the means arguments. I really don’t have a problem with people that are attempting to use EA to decide between multiple good outcomes.
I’m definitely not engaged enough with the Effective Altrusits to know where the plurality of thought lies, so I was trying to respond in the context of this argument being put forward on behalf of Effective Altruists.
The only part I’d say applies to all EA, is the brand taint that SBF has done in the public perception.
The speed doesn't really matter if their end goal is morally wrong. A slower speed might give them an advantage to not overshoot and get backlash or it gives artists and the public more time to fight back against EA, but it doesn't hide their ill intentions.
There's a fair amount of EA discussion of utilitarianism's problems. Here's EA founder Toby Ord on utilitarianism and why he ultimately doesn't endorse it:
https://forum.effectivealtruism.org/posts/YrXZ3pRvFuH8SJaay/...
Very few in the EA community want to speed AI adoption. It's far more common to think that current AI companies are being reckless, and we need some sort of AI pause so we can do more research and ensure that AI systems are reliably beneficial.
The all-time most upvoted post on the EA Forum condemns SBF: https://forum.effectivealtruism.org/allPosts?sortedBy=top&ti...
I’ve had to explain myself a few times on this, so clearly I communicated badly.
I probably should have said _those_ Effective Altruists are shitty utilitarians. I was attempting—and since I’ve had to clarify a few times clearly failed—to take aim at the effective altruists that would make the utilitarian trade off that the commenter mentioned.
In fact, there’s a paragraph from the Toby Ord blog post that I wholeheartedly endorse and I think rebuts the exact claim that was put forward that I was responding to.
So, my words were too broad. I don’t actually mean all effective altruists are shitty utilitarians. But the ones that would make the arguments I was responding to are.
I think Ord is a really smart guy, and has worked hard to put some awesome ideas out into the world. I think many others (and again, certainly not all) have interpreted and run with it as a framework for shitty utilitarianism.
The central contention of Effective Altruism, at least in practice if not in principle, seems to be that the value of thinking, feeling persons can be and should be reduced to numbers and objects that you can do calculations on.
Maybe there's a way to do that right. I suppose like any other philosophy, it ends up reflecting the personalities and intentions of the individuals which are attracted to and end up adopting it. Are they actually motivated by identifying with and wanting to help other people most effectively? Or are they just incentivized to try to get rid of pesky deontological and virtue-based constraints like empathy and universal rights?
Plus, describing this as "speed the rate of life saving ai technology in the hands of everyone" is… A Reach.
It didn't require voice theft, they could have easily found a volunteer or paid for someone else.
Agree, and it'd be more authentic
Are we surprised by this bankruptcy. As neat as AI is, it is only a thing because the corporate class see it as a way to reduce margins by replacing people with it. The whole concept is bankrupt.
I don’t think any said anything about being surprised by it?
Problem is they really believe we either can't tell the difference between a human and an AI model eventually, or they think we don't care. Don't they understand the meaning of art?
100% this.
It’s shocking to me how people cannot see this.
The only surprise here is that they didn’t think she’d push back. That is what completes the multilayered cosmic and dramatic irony of this whole vignette. Honestly feels like Shakespeare or Arthur Miller might have written it.
That is wrong on several levels. First off, it ignores the role of massive researchers. Should we start de-automating processes to employee more people at the cost of worsening margins?
And they could have totally get away with it by never mentioning the name of Scarlett. But of course, that is not what they wanted.
Edit: to clarify, since it is not exactly identical voice, or even not that close, they can plausibly deny it, and we never new what their intention was.
But in this case, they have clearly created the voice to represent Scarlett's voice to demonstrate the capabilities of their product in order to get marketing power.
When studios approach an actress A and she refuses, then another actress B takes the role, is that infringing on A's rights? Or should they just scrap the movie?
Maybe if they replicated a scene from the A's movies or there was striking likeness between the voices... but not generally.
The scenario would have been that they approach none.
Normally I'd agree if this were some vague "artist style", but this was clearly an attempt to duplicate a living person, a media celebrity no less.
Is this different from the various videos of the Harry Potter actors doing comedic high fashion ads? Because those were very well received.
https://www.youtube.com/watch?v=ipuqLy87-3A
I think anti-deepfake legislation needs to consider fair use, especially when it comes to parody or other commentary on public figures. OpenAI's actions do not qualify as fair use.
The problem with that idea is that I can hide behind it while making videos of famous politicians doing really morally questionable things and distributing them on YouTube. The reason Fair Use works with regular parodies in my opinion is that everyone can tell that it is obviously fake. For example, Saturday Night Live routinely makes joking parody videos of elected officials doing things we think might be consistent with their character. And in those cases it's obvious that it's being portrayed by an actor and therefore a parody. If you use someone's likeness directly I think that it must never be fair use or we will quickly end up in a world where no video can be trusted.
Let me start by saying I despise generative AI and I think most AI companies are basically crooked.
I thought about your comment for a while, and I agree that there is a fine line between "realistic parody" and "intentional deception" that makes deepfake AI almost impossible to defend. In particular I agree with your distinction:
- In matters involving human actors, human-created animations, etc, there should be great deference to the human impersonators, particularly when it involves notable public figures. One major difference is that, since it's virtually impossible for humans to precisely impersonate or draw one another, there is an element of caricature and artistic choice with highly "realistic" impersonations.
- AI should be held to a higher standard because it involves almost no human expression, and it can easily create mathematically-perfect impersonations which are engineered to fool people. The point of my comment is that fair use is a thin sliver of what you can do with the tech, but it shouldn't be stamped out entirely.
I am really thinking of, say, the Joe Rogan / Donald Trump comedic deepfakes. It might be fine under American constitutional law to say that those things must be made so that AI Rogan / AI Trump always refer to each other in those ways, to make it very clear to listeners. It is a distinctly non-libertarian solution, but it could be "necessary and proper" because of the threat to our social and political knowledge. But as a general principle, those comedic deepkfakes are works of human political expression, aided by a fairly simple computer program that any CS graduate can understand, assuming they earned their degree honestly and are willing to do some math. It is constitutionally icky (legal term) to go after those people too harshly.
I think that as long as a clear "mark of parody" is maintained such that a reasonable person could distinguish between the two, AI parodies are probably fine. The murkiness in my mind is expressly in the situation in the first episode of Black Mirror, where nobody could distinguish between the AI-generated video and the prime minister actually performing the act. Clearly that is not a parody, even if some people might find the situation humorous. But if we're not careful we give people making fake videos cover to hide behind fair use for parody.
I think you and I have the same concerns about balancing damage to the societal fabric against protecting honest speech.
I’m guessing you’re referring to people still thinking Sarah Palin said she could see Russia from her house, that was from a SNL skit and an amazing impression from Tina Fey. I agree, people have a hard time separating reality from obvious parody, how could we expect them to make a distinction with intentional imitation. Society must draw a clear line that it is not ok to do this.
That has a much better chance of falling under fair use (parody, non-commercial) if the actors ever tried to sue.
There is a major difference between parodying someone by imitating them while clearly and almost explicitly being an imitation; and deceptively imitating someone to suggest they are associated with your product in a serious manner.
Those are parodies and not meant at any point for you to believe the actual Harry Potter actors were involved.
One is a company with a nearly $100 billion valuation using someone's likeness for their own commercial purposes in a large-scale consumer product, which consumers would plausibly interpret as a paid endorsement, while the other seems to be an amateur hobbyist nobody has ever heard of making a parody demo as an art project, in a way that makes it clear that the original actors had nothing to do with it. The context seems pretty wildly different to me.
I'm guessing if any of the Harry Potter actors threatened the hobbyist with legal action the video would likely come down, though I doubt they would bother even if they didn't care for the video.
There’s a big difference between a one off replica and person-as-a-service.
Is a billion dollar AI company utilizing someone's voice against their will in a flagship product after they said no twice different from a random Youtube channel making comedy videos?
I think so but that could just be me.
Why do you have an issue with them taking someone's likeness to use in their product but not with them taking someone's work to use in their product?
Because this isn't training an audio model along with a million other voices to understand English, etc. It's clearly meant to sound exactly like that one celebrity.
I suspect a video avatar service that looked exactly like her would fall afoul of fair use as well. Though an image gen that used some images of her (and many others) to train and spit out generic "attractive blonde woman" is fair use in my opinion.
Chances are this is. Basically same as LoRA. One of go-to tools for these literally uses Diffusion model and work on spectrograms as images.
Okay so as long as we steal enough stuff then it's legal.
An actress that specifically played the voice of AI in a movie about AI no less.
Sure they could have taken her to court but right now they don't want the bad publicity, especially since it would put everything else in the shadow of such a scandalous "story". Better to just back off, let S.J. win and move on and start planning on they're gonna spend all that paper money they got with announcement of a new, more advanced model. It's a financial decision and a fairly predictable one. I'm glad she won this time.
Paper money from the model they're giving away for free?
I mean if you don't think these kinds of positive announcements don't increase the value of the company or parent company then I don't really know how to convince you as it's a standard business principle.
There isn’t a positive announcement here, what is wrong with you?
This reads like “we got caught red handed” and doing the bare minimum for it to not appear malicious and deliberate when the timeline is read out in court.
I believe there's a difference between building a sustainable and profitable business and pumping the stock.
She also won big against Disney. They backed down even though it appeared the contract was on their side. Iger apologized.
https://www.bbc.co.uk/news/business-58757748.amp
Probably (and rightfully) feared that, had Disney stuck with their position, other MCU actors would be much, much harsher in new contract negotiations - or that some would go as far and say "nope, I quit".
If the purpose is to trade on the celebrity voice and perceived association, and its subject to California right of personality law, then, yes, we're saying that that has been established law for decades.
That's not the purpose though, clearly. If anything, you could make the argument that they're trading in on the association to the movie "Her", that's it. Neither Sky nor the new voice model sound particularly like ScarJo, unless you want to imply that her identity rights extend over 40% of all female voice types. People made the association because her voice was used in a movie that features a highly emotive voice assistant reminiscent of GPT-4o, which sama and others joked about.
I mean, why not actually compare the voices before forming an opinion?
https://www.youtube.com/watch?v=SamGnUqaOfU
https://www.youtube.com/watch?v=vgYi3Wr7v_g
-----
https://www.youtube.com/watch?v=iF9mrI9yoBU
https://www.youtube.com/watch?v=GV01B5kVsC0
Whether you think it sounds like her or not is a matter of opinion, I guess. I can see the resemblance, and I can also see the resemblance to Jennifer Lawrence and others.
What Johannson is alleging goes beyond this, though. She is alleging that Altman (or his team) reached out to her (or her team) to lend her voice, she was not interested, and then she was asked again just two days before GPT-4o's announcement, and she rejected again. Now there's a voice that, in her opinion, sounds a lot like her.
Luckily, the legal system is far more nuanced than just listening to a few voices and comparing it mentally to other voices individuals have heard over the years. They'll be able to figure out, as part of discovery, what lead to the Sky voice sounding the way it does (intentionally using Johannson's likeness? coincidence? directly trained off her interviews/movies?), whether OpenAI were willing to slap Johannson's name onto the existing Sky during the presentation, whether the "her" tweet and the combination of the Sky voice was supposed to draw the subtle connection... This allegation is just the beginning.
I honestly don't think it is a matter of opinion, though. Her voice has a few very distinct characteristics, the most significant of which being the vocal fry / huskiness, that aren't present at all in either of the Sky models.
Asking for her vocal likeness is completely in line with just wanting the association with "Her" and the big PR hit that would come along with that. They developed voice models on two different occasions and hoped twice that Johannson would allow them to make that connection. Neither time did she accept, and neither time did they release a model that sounded like her. The two day run-up isn't suspicious either, because we're talking about a general audio2audio transformer here. They could likely fine-tune it (if even that is necessary) on her voice in hours.
I don't think we're going to see this going to court. OpenAI simply has nothing to gain by fighting it. It would likely sour their relation to a bunch of media big-wigs and cause them bad press for years to come. Why bother when they can simply disable Sky until the new voice mode releases, allowing them to generate a million variations of highly-expressive female voices?
I haven’t hear the GPT-4o voice before. Comparing the video to the video of Johansson’s voice in “her”, it sounds pretty similar. Johansson’s performance there sounds pretty different from her normal speaking voice in the interview - more intentional emotional inflection, bubbliness, generally higher pitch. The GPT-4o voice sounds a lot like it.
From elsewhere in the thread, likeness rights apparently do extend to intentionally using lookalikes / soundalikes to create the appearance of endorsement or association.
We can seriously say that, yes. The courts have been saying this in the US for over 30 years. See Midler v. Ford Motor Co.
Tom Waits won a lawsuit against Doritos too.
It could be trained on Scarlett's voice though, there's plenty of recorded samples for OpenAI to use. It's pretty damning for them to take down the voice right away like that
Her statement claims the voice was taken down at her attorney's insistence.
I think the copyright industry wants to grab new powers to counter the infinite capacity of AI to create variations. But that move would knee cap the creative industry first, newcomers have no place in a fully copyrighted space.
It reminds me of how NIMBY blocks construction to keep up the prices. Will all copyright space become operated on NIMBY logic?
I think we should all be held to the standard of “Weird” Al Yankovic. In personal matters consent is important.
this is correct. in fact the fcc has already clarified this for the case of robocalls. https://www.fcc.gov/document/fcc-makes-ai-generated-voices-r...
I thought it sounded like Jodie Foster.
Scar Jo thought it sounded like herself, and so did people who knew her personally.
That is what matters. OWNERSHIP over her contributions to the world.
I mostly agree with you, but I actually don't think it matters if it sounded exactly like her or not. The crime is in the training: did they use her voice or not?
If someone licenses an impersonator's voice and it gets very close to the real thing, that feels like an impossible situation for a court to settle and it should probably just be legal (if repugnant).
If OpenAI commissioned a voice actor to lend their voice to the Sky model, and cast on the basis of trying to get someone who is similar sounding to the Scarlett Johannson, but then did not advertise or otherwise use the voice model created to claim it was Scarlett Johannson - then they're completely in the clear.
Because then the actual case would be fairly bizarre: an entirely separate person, selling the rights to their own likeness as they are entitled to do, is being prohibited from doing that by the courts because they sound too much like an already famous person.
EDIT: Also up front I'm not sure you can entirely discuss timelines for changing out technology here. We have voice cloning systems that can do it with as little as 15 seconds of audio. So having a demo reel of what they wanted to do that they could've used on a few days notice isn't unrealistic - and training a model and not using it or releasing it also isn't illegal.
That's confidently incorrect. Many others already posted that this has been settled case law for many years. I mean would you argue that if someone build a macbook lookalike, but not using the same components would be completely clear?
I ask you what do you call the Framework [1]? Or Dell's offerings?[2] Compared to the Macbook? [3]
Look kind of similar right? Lot of familiar styling queues? What would take it from "similar" to actual infringement? Well if you slapped an Apple Logo on there, that would do it. Did OpenAI make an actual claim? Did they actually use Scarlett Johannson's public image and voice as sampling for the system?
[1] https://images.prismic.io/frameworkmarketplace/25c9a15f-4374...
[2] https://i.dell.com/is/image/DellContent/content/dam/ss2/prod...
[3] https://cdn.arstechnica.net/wp-content/uploads/2023/06/IMG_1...
You're not arguing your way out of jurisprudence, especially when the subject is a human and not a device nor IP. They (OpenAI) fucked up.
There is not clear jurisprudence on this. They're only in trouble if they actually used ScarJo's voice samples to train the model, or if they intentionally tried to portray their imitation as her without her permission.
The biggest problem on that front (assuming the former is not true) is Altman's tweets, but court-wise that's defensible (though I retract what I had here previously - probably not easily) as a reference to the general concept of the movie.
Because otherwise the situation you have is OpenAI seeking a particular style, hiring someone who can provide it, not trying to pass it off as that person (give or take the Tweet's) and the intended result effectively being: "random voice actress, you sound too much like an already rich and famous person. Good luck having no more work in your profession" - which would be the actual outcome.
The question entirely hinges on, did they include any data at all which includes ScarJo's voice samples in the training. And also whether it actually does sound similar enough - Frito-Lay went down because of intent and similarity. There's the hilarious outcome here that the act of trying to contact ScarJo is the actual problem they had.
EDIT 2: Of note also - to have a case, they actually have to show reputational harm. Of course on that front, the entire problem might also be Altman. Continuing the trend I suppose of billionaires not shutting up on Twitter being the main source of their legal issues.
Are you a lawyer?
Grey laptops that share some ideas in their outline while being distinct enough to not get lawyers from Cupertino on their necks?
Well Sam Altman tweeted "her" so that does seem to me like they're trying to claim a similarity to Scarlett Johannson.
It is not an impossible situation, courts have settled it, and what you describe is not how the law works (despite how many computer engineers think to the contrary.)
Courts have settled almost nothing related to AI. We don't even know if training AI using copyrighted works is a violating of copyright law.
Please point to a case where someone was successfully sued for sounding too much like a celebrity (while not using the celebrity's name or claiming to be them).
Multiple cases already answering your question in this thread.
Midler vs Ford: https://en.wikipedia.org/wiki/Midler_v._Ford_Motor_Co.
This has been settled law for 34 years. See Tom Waits v Frito-Lay.
They literally hired an impersonator, and it cost them 2.5 million (~6 million today).
https://www.latimes.com/archives/la-xpm-1990-05-09-me-238-st...
That case seems completely dissimilar to what OpenAI did.
Frito-Lay copied a song by Waits (with different lyrics) and had an impersonator sing it. Witnesses testified they thought Waits had sung the song.
If OpenAI were to anonymously copy someone's voice by training AI on an imitation, you wouldn't have:
- a recognizable singing voice
- music identified with a singer
- market confusion about whose voice it is (since it's novel audio coming from a machine)
I don't think any of this is ethical and think voice-cloning should be entirely illegal, but I also don't think we have good precedents for most AI issues.
Let me connect the dots for you.
Company identifies celebrity voice they want. (Frito=Waits, OpenAi=ScarJo)
Company comes up with novel thing for the the voice to say. (Frito=Song, OpenAI=ChatGpt)
Company decides they don’t need the celebrity they want (Frito=Waits, OpenAI=ScarJo) and instead hire an impersonator (Frito=singer, {OpenAI=impersonator or OpenAI=ScarJo-public-recordings}) to get what they want (Frito=a-facsimile-of-Tom-Waitte’s-voice-in-a-commercial, OpenAi=a-fascimilie-of-ScarJo’s-voice-in-their-chatbot)
When made public, people confuse the fascimilie as the real thing.
I don’t see how you don’t see a parallel. It’s literally best for beat the same, particularly around the part about using an impersonator as an excuse.
As I understand it (though I may be wrong) in music sampling cases, it doesn’t matter if the “sample” is using an actual clip from a recording or if were recreated from scratch using a new media (e.g. direct midi sequence), if a song sampling another song is recognizable it is still infringing.
Sampling is not the same as duplication. Sampling is allowed as it's a derivitive work as long as it's substantially different from the original.
It's a "I know it when I see it" situation so it's not clear cut.
Oh, the day when an artist could sample other artists without attribution and royalties is long gone. The music labels are very hard on this these days.
https://en.wikipedia.org/wiki/Personality_rights
This is a civil issue, and actors get broad rights to their likeliness. Kim Kardashian sued Old Navy for using a look-alike actress in an ad; old Navy chose to settle, which makes it appear like "the real actress wasn't involved in any way" may not be a perfect defense. The timeline makes it clear they wanted it to sound like Scarlett's voice, the actual mechanics on how they got the AI to sound like that is only part of the story.
Does that mean if cosplayers dress up like some other character, they can use that version of the character in their games/media? I think it should be equally simple to settle. It's different if it's their natural voice. Even then, it brings into question whether they can use "doppelgangers" legally.
She doesn't own most (all probably) of her contributions to the world.
If the voice was only trained on the voice of the character she played in Her, would she have any standing in claiming some kind of infringement?
IANAL, but – I think it's most likely to be an infringement on her right of publicity (i.e. the right to control the commercial value of your name, likeness, etc.)
She doesn't have to own anything to claim this right, if the value of her voice is recognizable.
More notably for legal purposes, there were several independent news reports corroborating the vocal similarity.
...and sama's tweet referencing "Her"
I'm not sure how much you currently legally own imitations of your own voice. There's a whole market for voice actors who can imitate particular famous voices.
Should have renamed it
https://en.wikipedia.org/wiki/Sosumi
Or
https://www.reddit.com/r/todayilearned/comments/9n44b6/til_t...
Clearly Sam Altman though it sounded like ScarJo as well :-(
At least in past court cases I’m familiar, you can’t use an impersonator and get people to think it’s the real thing.
It’s not like Tom Waits ever wanted to hock chips
https://www.latimes.com/archives/la-xpm-1990-05-09-me-238-st...
but did openAI make any claims about whose voice this is? Just because a voice sounds similar or familiar, doesn't mean it's fraudulent.
Just read the top post of the thread you're responding to-
To me reference sounds more to towards omni than her voice
What’s “omni”?
GTP-4o is the new model, the o stands for Omni.
That’s not a gamble they are willing to take to court of law or public opinion.
Or Bette Midler singing for ford. She turned them down. They used a sound alike, she sued and won
https://en.m.wikipedia.org/wiki/Midler_v._Ford_Motor_Co.
They used a sound-alike and had her sing one of her songs. I believe that's a different precedent, in that it's leveraging her fame.
Imo Sky's voice is distinct enough from Scarlett, and it wasn't implied to _be_ her.
Sam's "Her" tweet could be interpreted as such, but defending the tweet as the concept of "Her", rather than the voice itself, is.
Does not the 'her' tweet give away the game. Aas you said, it was a midler impersator singing one of midlers songs. In this case they have a voice for their AI assistant/phone sex toy that is very much like the actress that played a famous ai assistant/phone sex toy. Even if he is taken as meaning the concept it's very very damning. If they had, instead, mimic'ed another famous actor's voice that hasn't played a robot/ai/whatever and used that would that really be any better though? Christopher Walken, say, or hell Bette Midler?
There’s a nice YouTube doc telling the story of this, and Tom Waits’ hatred of advertising - https://youtu.be/W7J01e-OIMA?si=57IJooNwg5oTfh62
If they really hired someone who sounds just like her it's fair game IMO. Johanssen can't own the right to a similar voice just like many people can have the same name. I think if there really was another actress and she just happens to sound like her, then it's really ok. And no I'm not a fan of Altman (especially his worldcoin which I view as a privacy disaster)
I mean, imagine if I happened to have a similar voice to a famous actor, would that mean that I couldn't work as a voice actor without getting their OK just because they happen to be more famous? That would be ridiculous. Pretending to be them would be wrong, yes.
If they hired someone to change their voice to match hers, that'd be bad. Yeah. If they actually just AI-cloned her voice that's totally not OK. Also any references to the movies. Bad.
But clearly they are advertising as her (no pun intended), which is a gray area.
Yeah that was the bad part. Agreed there.
I wonder if they deliberately steered towards this for more marketing buzz?
Discovery process will be interesting
Why are you mad? We have no rights to the sound of our voice. There is nothing wrong with someone or something else making sounds that sound like us, even if we don’t want it to happen.
No one is harmed.
The law can actually be interesting and nuanced on this: http://law2.umkc.edu/faculty/projects/ftrials/communications...
I think it's a different argument with respect to famous media celebrities* too.
If someone clones a random person's voice for commercial purposes, the public likely has no idea who the voice's identity is. Consequently, it's just the acoustic voice.
If someone clones a famous media celebrity's voice, the public has a much greater chance of recognizing the voice and associating it with a specific person.
Which then opens a different question of 'Is the commercial use of the voice appropriating the real person's fame for their own gain?'
Add in the facts that media celebrities' values are partially defined by how people see them, and that they are often paid for their endorsements, and it's a much clearer case that (a) the use potentially influenced the value of their public image & (b) the use was theft, because it was taking something which otherwise would have had value.
Neither consideration exists with 'random person's voice' (with deference to voice actors).
* Defined as 'someone for whom there is an expectation that the general public would recognize their voice or image'
Are you sure? You certainly have rights to your likeness--it can't be used commercially without permission. Di you know this doesn't cover your voice?
Everyone is so mad about them stealing a beloved celebrity’s voice. What about the millions of authors and other creators whose copyrighted works they stole to create works that resemble and replace those people? Not famous enough to generate the same outrage?
Welcome to the world where the "fuck the creatives" brigade wants everything for free.
A mentality that devalues creative work
I think the unique thing about this case is not specifically the "voice theft", but that OpenAI specifically asked for permission and were denied, which eliminates most of the usual plausible deniability that gets trotted out in these cases.
Same here and that voice really was the only good one. I don't know why they don't bring the voices from their API over, which are all much better, like Nova or Shimmer (https://platform.openai.com/docs/guides/text-to-speech)
I think because it is not text-to-speech. It probably isn't simple to transfer.
The previous chatgpt voice mode uses text to speech, and has different voices than the OpenAI API. Seemed weird to me too.
I can still access the sky voice even though it is supposed to be "yanked".
There’s still a Sky option but the actual voice has been changed.
https://x.com/jam3scampbell/status/1791338109709287511
I had to go look at what voice I picked once I heard the news, it was Sky. I listened to them all and thought it sounded the best. I didn’t make any connection to her (Scar Jo or the movie) when going through the voices, but I wasn’t listening for either. I don’t think I know her voice well enough to pick it out of a group like that.
Maybe I liked it best because it felt familiar, even if I didn’t know why. I’m a bit disappointed now that she didn’t sign on officially, but my guess is that Altman just burned his bridge to half of Hollywood if he is looking for a plan B.
I wouldn't necessarily call that damning. "Soundalikes" are very common in the ad industry.
For example, a car company approached the band sigur ros to include some of their music in a car commercial. Sigur ros declined. A few months later the commercial airs with a song that sounds like an unreleased sigur ros song, but really they just paid a composer to make something that sounds like sigur ros, but isn't. So maybe openai just had a random lady with a voice similar to Scarlett so the recording.
Taking down the voice could just be concern for bad press, or trying to avoid lawsuits regardless of whether you think you are in the right or not. Per this* CNN article:
So, Johansson's lawyers probably said something like "I'll sue your pants off if you don't take it down". And then they took it down. You can't use that as evidence that they are guilty. It could just as easily be the case that they didn't want to go to court over this even if they thought they were legally above board.
* https://www.cnn.com/2024/05/20/tech/openai-pausing-flirty-ch...
Case law says no.
There have been several legal cases where bands have sued advertisers for copying their distinct sound. Here are a few examples:
The Beatles vs. Nike (1987): The Beatles' company, Apple Corps, sued Nike and Capitol Records for using the song "Revolution" in a commercial without their permission. The case was settled out of court.
Tom Waits vs. Frito-Lay (1988): Tom Waits sued Frito-Lay for using a sound-alike in a commercial for their Doritos chips. Waits won the case, emphasizing the protection of his distinct voice and style.
Bette Midler vs. Ford Motor Company (1988): Although not a band, Bette Midler successfully sued Ford for using a sound-alike to imitate her voice in a commercial. The court ruled in her favor, recognizing the uniqueness of her voice.
The Black Keys vs. Pizza Hut and Home Depot (2012): The Black Keys sued both companies for using music in their advertisements that sounded remarkably similar to their songs. The cases were settled out of court.
Beastie Boys vs. Monster Energy (2014): The Beastie Boys sued Monster Energy for using their music in a promotional video without permission. The court awarded the band $1.7 million in damages.
Disregarding the cases settled out of court (which have nothing to do with case law):
1) Tom Waits vs Frito-Lay: Frito-Lay not only used a soundalike to Tom Waits, but the song they created was extremely reminiscent of "Step Right Up" by Waits.
2) Bette Midler vs. Ford Motor Company: Same thing - this time Ford literally had a very Midler-esque singer sing an exact Midler song.
3) Beastie Boys vs. Monster Energy: Monster literally used the Beastie Boys' music, because someone said "Dope!" when watching the ad and someone at Monster took that to mean "Yes you can use our music in the ad".
Does Scarlett Johansson have a distinct enough voice that she is instantly recognizable? Maybe, but, well, not to me. I had no clue the voice was supposed to be Scarlett's, and I think a lot of people who heard it also didn't think so either.
It absolutely is if you've seen /Her/. It even nails her character's borderline-flirty cadence and tone in the film.
Actually until recently I thought the voice actor for "her" was Rashida Jones
I definitely thought "Sky" was Rashida Jones. I still do.
I've seen Her and the similarity of the voice didn't occur to me until I read about it. I guess it wasn't super distinct in the movie. Maybe if they'd had Christopher Walken or Shakira or someone with a really distinctive sound it would have been more memorable and noticable to me.
The Midler v. Ford decision said her voice was distinctive. Not the song.
And in an interesting coincidence: ScarJo recorded a Tom Waits cover album in 2008
Sucks that he had to do it, but the notion of Tom Waits making Rain Dogs and then pivoting to spending a bunch of time thinking about Doritos must be one of the funnier quirks of music history.
Sorry, that's apples-to-pizzas comparison. You're conflating work and identity.
There's an ocean of difference between mimicking the style of someone's art in an original work, and literally cloning someone's likeness for marketing/business reasons.
You can hire someone to make art in the style of Taylor Swift, that's OK.
You can't start selling Taylor Swift figurines by the same principle.
What Sam Altman did, figuratively, was giving out free T-Shirts featuring a face that is recognized as Taylor Swift by anyone who knows her.
But they aren't doing anything with her voice(allegedly?). They're doing something with a voice that some claim sounds like hers.
But if it isn't, then it is more like selling a figurine called Sally that happens to look a lot like Taylor Swift. Sally has a right to exist even if she happens to look like Taylor Swift.
Has there ever been an up and coming artist who was not allowed to sell their own songs, because they happened to sound a lot like an already famous artist? I doubt it.
TL;DR: This question had already been settled in 2001 [3]:
The court determined that Midler should be compensated for the misappropriation of her voice, holding that, when "a distinctive voice of a professional singer is widely known and is deliberately imitated in order to sell a product, the sellers have appropriated what is not theirs and have committed a tort in California."
I hope there's going to be no further hypotheticals after this.
-----
Yes, that's what a likeness is.
If you start using your own paintings of Taylor Swift in a product without her permission, you'll run afoul of the law, even though your painting is obviously not the actual Taylor Swift, and you painted it from memory.
Sally has a right to exist, not the right to be distributed, sold, and otherwise used for commercial gain without Taylor Swift's permission.
California Civil Code Section 3344(a) states:
Any person who knowingly uses another’s name, voice, signature, photograph, or likeness, in any manner, on or in products, merchandise, or goods, or for purposes of advertising or selling, or soliciting purchases of, products, merchandise, goods or services, without such person’s prior consent, or, in the case of a minor, the prior consent of his parent or legal guardian, shall be liable for any damages sustained by the person or persons injured as a result thereof.
Note the word "likeness".
Read more at [1] on Common Law protections of identity.
Wrong question.
Can you give me an example of an artist which was allowed to do a close-enough impersonation without explicit approval?
No? Well, now you know a good reason for that.
Tribute bands are legally in the grey area[2], for that matter.
[1] https://www.dmlp.org/legal-guide/california-right-publicity-...
[2] https://lawyerdrummer.com/2020/01/are-tribute-acts-actually-...
[3] https://repository.law.miami.edu/cgi/viewcontent.cgi?article...
The detail you're missing is that who claim it sounds like "her" includes the CEO of the company
The damning part is that they tried to contact her and get her to reconsider their offer only 2 days before the model was demoed. That tells you that at the very least they either felt a moral or legal obligation to get her to agree with their release of the model.
Or, they wanted to be able to say, yes, that is "her" talking to you.
I have no idea if they really used her voice, or it is a voice that just sounds like her to some. I'm just saying openai's behavior isn't a smoking gun.
If this isn't a smoking gun, I don't know what it.
I think people forget the last part of the definition, though. A Smoking gun is about as close as you get without having objective, non-doctored footage of the act. There's a small chance the gun is a red herring, but it's still suspicious.
As are disclaimers that celebrity voices are impersonated when there is additional context which makes it likely that the voice would be considered something other than a mere soundalike, like direct reference to a work in which the impersonated celebrity was involved as part of the same publicity campaign.
And liability for commercial voice appropriation, even by impersonation, is established law in some jurisdictions, including California.
The most famous case of voice appropriation was Midler vs Ford, which involved Ford paying a Midler impersonator to perform a well known Midler song, creating the impression that it was actually Bette.
Where are the signs or symbols tying Scarlett to the openAI voice? I don't think a single word, contextless message on a separate platform that 99% of openAI users will not see is significant enough to form that connection in users heads.
The Midler v. Ford decision said her voice was distinctive. Not the song.
The replies to Altman's message showed readers did connect it to the film. And people noticed the voice sounded like Scarlett Johansson and connected it to the film when OpenAI introduced it in September.[1]
How do you believe Altman intended people to interpret his message?
[1] https://www.reddit.com/r/ChatGPT/comments/177v8wz/i_have_a_r...
When people cheat on (relatively) small things, it's usually an indication they'll cheat on big things too
Stealing someone's identity is indeed one of those "big things".
Impersonating someone’s voice isn’t stealing anything, and certainly not their identity.
If they are a celebrity actor it sure is.
Voice actors especially. Voice acting isn't what ScarJo is really "known for" but she's done a ton of work for animated films so her distinct voice really is a part of her livelihood.
30+ year old established case precedent disagrees with you:
http://law2.umkc.edu/faculty/projects/ftrials/communications...
https://casetext.com/case/waits-v-frito-lay-inc
How did they even cheat here?
OpenAI did nothing wrong.
The movie industry does the same thing all the time. If an actor/actress says no they you find someone else who can play the same role.
Nothing? If you're acting like the sea witch in "The Little Mermaid" you're probably doing something wrong.
Key difference here is that Scarlett still has her voice.
I don't think that's quite the same. Are they going out and hiring impersonators of the actors who declined the role or digitally enhancing the substitute to look like them? That seems closer to what happened here.
If they are so "Open" they should reveal their training data which created this voice. I am sure it is just movie audio from S. Johansson's movies.
Which is what makes me wonder if this might grow into a galvanizing event for the pro-creator protests against these AI models and companies. What happened here isn't particularly unique to voices or even Scarlett Johansson, it is just how these companies and their products operate in general.
I think the only way for these protests to get really tangible results is in case we reach a ceiling in LLM capabilities. The technology in its current trajectory is simply too valuable both in economic and military applications to pull out of, and "overregulation" can be easily swatted citing national security concerns in regards to China. As far as I know, China has significantly stricter data and privacy regulations than the US when it comes to the private sector, but these probably count for little when it comes to the PLA.
We have almost run out of training data already so I’m not convinced they will get massively more generalised suddenly. If you give them reasoning tasks they haven’t seen before LLMs absolutely fall apart and produce essentially gibberish. They are currently search engines that give you one extremely good result that you can refine up to a point, they are not thinking even though there’s a little bit more understanding than search engines of the past.
I would love to see the providence of their training data.
I think you want the word “provenance.”
We need laws where companies are forced to reveal source of personal data. Like how did XYZ company get my contact info to spam me?
Who cheated whom? Out of what?
OpenAI only hires and is built on the culture that data and copyright is somehow free for the taking, otherwise they would have zero ways to make a profit or “build agi”
Given the timeline, I’m still baffled Sam Altman tweeted “her.” That just makes plausible deniability go away for a random shitpost.
I thought it was about functionality more than the specific voice.
I suspect that'll be OpenAI's defense.
That is what discovery is for. If this would ever get to that phase.
Someone from OpenAi hired the agency who hired the voice talent (or talents) for the voice data. They sent them a brief explaining what they are looking for, followed by a metric ton of correspondence over samples and contracts and such.
If anywhere during those written communications anyone wrote “we are looking for a ScarlettJ imitator”, or words to that effect, that is not good for OpenAI. Similarly if they were selecting between options and someone wrote that one sample is more Johansson than an other. Or if anyone at any point asked if they should clear the rights to the voice with Johansson.
Those are the discovery findings which can sink such a defense.
Discovery works both ways. The original Her voice actress[1] was recast to someone more SoCal in post-production, so there is evidence of the flirty erotic AI style itself not being a unique enough selling point.
It will come down to what makes the complaining celebrity's voice iconic, which for Scarjo is the 'gravelly' bit. Which smooth Sky had none of.
[1] actress reading poem: https://www.youtube.com/watch?v=eWEEAjRFJKc
Ok? What materials would you suspect discovery can uncover from Scarlett or her team?
Was recast to Scarlett Johansson. Hardly a good argument if you want to argue that her voice is not unique.
I argued Scarlett Johansson's natural voice is iconic and unique and it is not apparent in the Sky voice. The "new" intoned voice of Sky 4o is receiving most claims of likeness ripoff but this intoning part is not unique to ScarJo either since the Her production already had a version of it with the other VA. I doubt scarlett will suggest she was playing herself without creative direction from Her team.
Are you arguing that Her performance is not the resembling factor here but simply the natural voice of ScarJo at rest, disregarding Her?
I recall the basics from my contracts law class that it’s not against the law to hire an impersonator as long as you don’t claim it’s the celebrity.
So it’s legal to hire someone who sounds like SJ. And likely legal to create a model that sounds like her. But there will likely need to be some disclaimer saying it’s not her voice.
I expect that OpenAI’s defense will be something like “We wanted SJ. She said no, so we made a voice that sounded like her but wasn’t her.” It will be interesting to see what happens.
Odd. This wouldn't typically be covered in contracts class since it's an intellectual property issue beyond the scope of a contracts class. Scarlet Johansen didn't have a contract with OpenAI after all.
What you recall also doesn't sound correct given right of publicity laws.
Bette Midler and Tom Waits won cases where the companies didn't claim the impersonators were them.
What part of the functionality from the movie Her did you think it meant?
The same egomaniac tendencies that cause people like Elon Musk or Paul Graham to post the first dumbass thing that comes to their mind because they think everyone absolutely has to see how smart and witty they are.
Some people are just addicted to posting
Would love to see this get far enough for discovery to see how that all played out behind the scenes.
They’ll settle as soon as they figure that out. Idiot tax.
She has no incentive to settle and actually could win big by being the figurehead of the creative industry against AI. It’s understandable why she accepted a settlement from Disney, but there’s no reason why she should settle with a random startup that has no other influence on her employability in Hollywood.
OpenAI will settle, not sure how you read that in reverse.
Settling isn't unilateral. OpenAI can offer to settle, but if she doesn't accept, there will be no settlement.
I also said “they” instead of “her”, I’m confused as to why anybody misinterpreted what I said.
Everyone knows what you meant. But it’s not up to them to “settle”. If she brings forward a formal complaint they can offer to but she has no obligation or incentive to accept.
This may turn out to be something they can’t just buy their way out of with no other consequences.
And plenty of her peers have been fighting against AI content harvesting in their recent contract negotiations[1].
1. https://apnews.com/article/hollywood-ai-strike-wga-artificia...
She has a net worth of $100-$200m dollars. I doubt she wants $5-10m more.
OpenAI cannot hurt her standing in the industry— in fact, “ScarJo takes on Big Tech and wins”, in an era after the Hollywood unions called a strike and won protections from studios using generative AI for exactly this scenario, is ironically probably one of the best thing she can do for her image right now.
She is also one of the most litigious actresses in the industry, taking on Disney and winning what’s estimated to be 8 figures.
Good luck OpenAI!
- Two days before the GPT-4o launch, they contacted her agent and asked that she reconsider. (Two days! This means they already had everything they needed to ship the product with Scarlett’s cloned voice.)
New voice mode is a speech predicting transformer. "Voice Cloning" could be as simple as appending a sample of the voice to the context and instructing it to imitate it.
If they really did that then (A) it's not much better (B) they didn't even wait for an answer from Johansson (C) it's extraordinarily reckless to go from zero to big-launch feature in less than two days.
OP seems to be on the "They secretly trained on her voice" train. The only reason "Two Days!" would be damning is if a finetune was in order to replicate ScarJo's voice. In that sense, it's much better.
Open AI have launched nothing. There's no date for new voice mode other than, "alpha testing in the coming weeks to plus users". No-one has access to it yet.
I'm really confused by this claim. How is it that so many people tried this Sky voice if no one has access yet?
There is a voice mode in the GPT app that's been out (even for free users) for nearly a year now. There are a couple voices to chooses from and sky was one of them.
This mode works entirely differently from what Open AI demoed a few days ago (the new voice mode) but both seem to utilize the same base sky voice. All this uproar is from the demos of new sky which sounds like old sky but is a lot more emotive, laughs, a bit flirty etc.
Idk the voice in the 4o demo and the existing Sky voice seemed quite different to me. And Scarlett’s letter says she and her friends were shocked when they heard the 4o demo. This whole situation is about the newly unveiled voice in the demo. It’s different.
Sky voice is old
I know I have a reputation as an OpenAI hater and I understand why: it’s maybe 5-10% of the time that the news gives me the opportunity to express balance on this.
But I’ve defended them from unfair criticism on more than a few occasions and I feel that of all the things to land on them about this one is a fairly mundane screwup that could be a scrappy PM pushing their mandate that got corrected quickly.
The leadership for the most part scares the shit out of me, and clearly a house-cleaning is in order.
But of all the things to take them to task over? There’s legitimately damning shit this week, this feels like someone exceeded their mandate from the mid-level and legal walked it back.
It really doesn't sound like a "mid-level exceeding their mandate".
It sounds like Altman was personally involved in recruiting her. She said no and they took what they wanted anyway.
It feels weird to be defending Altman, but those of us who go hard on the legitimately serious shit are held to a high standard on being fair and while multiple sources have independently corroborated e.g. the plainly unethical and dubiously legal NDA shit Vox just reported on, his links to this incident seem thinly substantiated.
I’m not writing the guy a pass, he’s already been fired for what amount to ethics breaches in the last 12 months alone. Bad actor any way you look at it.
But I spent enough time in BigCo land to know stuff like this happens without the CEO’s signature.
I’d say focus on the stuff with clear documentary evidence and or credible first-hand evidence, there’s no shortage of that.
I get the sense this is part of an ambient backlash now that the damn is clearly breaking.
Of all the people who stand to be harmed by that leadership team, I think Ms. Johansson (of who I am a fan) is more than capable of seeing her rights and privileges defended without any help from this community.
There’s more evidence of Altman being personally involved in this incident than in him being personally involved in the OpenAI exit agreement, and he has denied the latter. I’m not sure I believe his denial in the latter case.
Having an NDA in exit terms you don’t get to see until you are leaving that claim ability to claw back your vested equity if you don’t agree seems more severely unethical, to be sure. But that doesn’t mean there’s more reason to blame it on Altman specifically. Or perhaps you take the stance that it reflects on OpenAI and their ethics whether or not Altman was personally involved, but then the same applies to the voice situation.
Scarlet Johansson literally mentioned that Sam personally reached out to her team.
The CTO was on stage presenting the thing and the CEO was tweeting about it.
Please explain for us which part of this is happening without the CEO's signature.
Of everyone who has been harmed and had their work stolen or copyright infringed by Sam's team, Scarlett Johansson is the one person (so far) who can actually force the issue and a change, and so the community is right to rally behind her because if they're so brazen about this, it paints a very clear picture of the disdain they hold the rest of us in.
You’re allegedly “not writing the guy a pass” but then you go on to do so anyways. If Johansson isn't lying and Altman did personally reach out to her, I really don’t see how you can even attempt to argue this is some middle manager gunning for a promotion. In the same way you’re complaining that she needs no help from the community in defending herself, Altman needs no help from you reaching this hard. Like how can you not see that hypocrisy?
If this isn't the thing that makes your blood boil, that's fine. The world could probably do with less boiling blood, and it is still early, more evidence may come out. However, she indicated in her statement that Altman asked her, not OpenAI. It seems credible that he would want to be involved.
Both sides of the story feel like we're slowly being brought to a boil: Sutskever's leaving feels like it was just a matter of time. His leaving causing a mess seems predictable. Perhaps I am numb to that story.
But stealing a large part of someone's identity after being explicitly told not to? This one act is not the end of the world, but feels like an acceleration down a path that I would rather avoid.
I'm beginning to think this Sam Altman guy isn't so trustworthy.
You beat me to it.
Bon mots apart, he really appears to have an innate capacity for betrayal.
Was there something in the water? Lots of rumbles around the early OpenAI members and questionable behavior.
And he must be a helluva pitchman, given the weird fired/hired debacle. Mixed with some of the resignations that made the HN front page recently, apparently anyone leaving OpenAI signed away the right to speak. I even find it odd that their statement says they hired a voice actress, but they want to protect her privacy? Seems like a helpful alibi if true, or likely, said actress has signed an agreement to never reveal she worked with OpenAI.
that took a while ;)
And perhaps not consistently candid either.
The voice was shipped last september.
Do you know if the voice was the same back in september?
Scarlet says she was shocked to hear the voice in the 4o demo, and they had requested her consent (for the 2nd time) 2 days prior to the demo. If that demo voice was the same as the existing Sky voice, this wouldn’t be happening.
It was essentially the same, the new one is definitely more human like. I have been calling it Scarjo since then.
https://x.com/sampullara/status/1735122897663094853The voices were available for some time for the ChatGPT TTS model, but it seems that they reused them for the 4o audio output, which sounds significantly more human-like. I’ve heard the Sky voice before and never made the connection. I did think of Johansson though during the live demo, as the voice + enhanced expressiveness made it sound much like the movie Her.
To each their own. I personally didn't get Scarlett Johansson vibes from the voice on the GPT-4o demo (https://openai.com/index/hello-gpt-4o/) even though I'm a huge fan of hers (loved Her, loved Jojo Rabbit, even loved Lucy, and many many others) and have watched those and others multiple times. I'd even say I have a bit of a celebrity crush.
To me it's about as close to her voice as saying "It's a woman's voice". Not to say all women sound alike but the sound I heard from that video above could maybe best be described and "generic peppy female American spokesperson voice"
Even listening to it now with the suggestion that it might sound like her I don't personally hear Scarlett Johansson's voice from the demo.
There may be some damming proof where they find they sampled her specifically but saying they negotiated and didn't come to an agreement is not proof that it's supposed to be her voice. Again, to me it just sounds like a generic voice. I've used the the version before GPT-4o and I never got the vibe it was Scarlett Johansson.
I did get the "Her" vibe but only because I was talking to a computer with a female voice and it was easy to imagine that something like "Her" was in the near future. I also imagined or wished that it was Majel Barrett from ST:TNG, if only because the computer on ST:TNG gave short and useful answers where as ChatGPT always gives long-winded repetitive annoying answers
OpenAI confirmed it by removing the voice immediately after Johansson's lawyers reached out
That's not confirmation. That's called prudence.
Ask to use Alice's likeness in my product, get declined. Ask again. Debut product, users say "this sounds a lot like Alice". The day my product releases, I make a public statement referencing a movie Alice starred in. When Alice asks how I made my product, I delete all access to the product.
How else am I supposed to interpret this?
"better safe than sorry"
- any legal team, ever, in the history of the universe
AFAIK they yanked it pretty quickly and the subsequent scandal has widely informed people that it was not authorized by Scarlett Johansson. So while it was clearly a violation resulting from a sequence of very stupid decisions by OpenAI, I am not sure if there would be much in the way of damages.
Johansson is rich. The real value she could get from this would be as an advocate for the rights of creatives, performers and rights holders in the face of AI. If this goes to discovery OpenAI is done.
How much do you think Disney or Universal Music or Google or NYT would give to peek inside OpenAI's training mixture to identify all the infringing content?
Not a lawyer, but as far as I understand; Certainly, the lack of significant damages would usually prevent a lawsuit because there wouldn't be any money in it, but since the amount of damages are figured out during the trial and the lawsuit alone would be damaging to OpenAI the threat of a lawsuit could be enough for OpenAI to try and settle quickly out of court. Though OpenAI would have to weigh that with against encouraging other similar lawsuits - they might think that dragging out the lawsuit might be better for them as they can continue establishing a foothold in AI in the interim even if they know they will lose and know that discovery would be damaging.
As you state, Mrs. Johansson would benefit reputationally form the lawsuit and could act as a front for other powerful institutions who would greatly benefit from the lawsuit.
In cases like this, don't damages essentially equate to the profit a company makes from the false association with the celebrity?
Otherwise, it'd be impossible to show damages if you weren't personally being denied business because of the association.
That is one way to measure damages but it's difficult. It would be easier to measure what she would have been paid if she had agreed to do it, Mrs. Johansson is in the business of selling rights to use her likeness for large amounts of money - it would be easy to demonstrate that she would have been paid at least as much as the use of her voice in the movie Her. Since she did decline the offer it would be easy to suggest that the amount she would have agreed to would have been much more than the movie Her.
The other aspect is that OpenAIs infringement is open ended, in that it is not equivalent to the single use within a single movie, it is substantially more. People who want Mrs. Johansson to voice their projects could now instead use OpenAI to obtain the same likeness - so the damages are for the deprivation all the future earnings Mrs. Johansson could have made.
Then there is reputational damage, if someone using OpenAI generated and disseminated a voice message before committing a heinous crime then the public would associate the voice with the crime and Mrs. Johansson would be intrinsically linked to the same crime. People hearing her voice would be reminded of the crime. This would again prevent Mrs. Johansson from making money from her likeness but would also negatively impact all aspects of her life and all future earnings, business dealings, and personal life.
Unusually the initial remedy is an injunction to prevent further infringement but OpenAI yanked it immediately so no injunction was needed. One way to repair reputational damage is to pay for news media to widely and publicly correct the reputationally damaging falsehood. As this has been a substantial scandal that media coverage has been already given for free and it is unlikely that there is anyone left who still believes that OpenAI is using Mrs. Johansson's likeness with either tacit or explicit permission.
As stated by your peer comment there are still reasons for Mrs. Johansson to bring a lawsuit for damages even where there are no significant damages - the lawsuit itself would be damaging to OpenAI so it would be in OpenAIs interest to pay to avoid it. At the same time it would be in other large companies interests for it to continue, so there may be a bidding war on whether or not Mrs. Johansson continues with the lawsuit. In addition it would benefit Mrs. Johansson personally and professionally to be seen as a champion of the rights of artists - especially at a time when AI companies like OpenAI are trampling all over those rights.
With the recent departures at OpenAI it seems that all ethics and morals are going down the drain and OpenAI becoming the big-bully.
There were never any. None of the models or code are actually open. It claims to be a nonprofit but is effectively a for profit company pulling the strings of a nonprofit just to avoid taxes.
I guess you are right. The era of "don't be evil" if there ever was one, is now long gone and forgotten.
This is a bit unfair. Some people left OpenAI on the ground of ethics, because they were unsatisfied with how this supposed nonprofit operates. The ethics was there, but OpenAI got rid of it.
ChatGPT is way better than to need stupid ripoffs than this
Sam should be ashamed to have ever thought of ripping off anyone's voice, let alone done it and rolled it out.
They are building some potentially world-changing technology, but cannot rise above being basically creepy rip-off artists. Einstein was right about requiring improved ethics to meet new challenges, and also that we are not meeting that requirement.
sad to see
It's part and parcel of the LLM field's usual disdain for any property rights that might belong to other people. What they did here is not categorically different than scraping every author and visual artist on the internet - but in this instance they've gone and brazenly "copied" (read: stolen) from one of the few folks with more media clout than they themselves have.
Of course it’s not. All of ChatGPT is a ripoff. That’s what training data (which they did not license) is.
It's also worth noting that Sam Altman admitted that he had only used GPT4o for one week before it was released. It's possible that in the rush to release before Google's IO event, they made the realisation of the likeness of the voice to Scarlett Johansen way too late hence the last minute contact with her agent.
https://www.youtube.com/watch?v=fMtbrKhXMWc
The "Sky" voice and it's likeness to SJo's have been there in the ChatGPT app for months.
Then asking Johansson for permission months before was pure coincidence?
Almost as if they knew that they cloned her voice without her permission.
Don't hear any arguments on how this is fair-use. (It isn't)
Why? Because everyone (including OpenAI) knows it clearly isn't fair-use even after pulling the voice.
Because it's a right of personality issue, not copyright, and there is no fair use exception (the definition of the tort is already limited to a subset of commercial use which makes the Constitutional limitation that fair use addresses in copyright not relevant, and there is no statutory fair use exception to liability under the right of personality.)
you're doing god's work; ignorance is never-ending
Could they have made it look less like Midler vs Ford?
"Midler was asked to sing a famous song of hers for the commercial and refused. Subsequently, the company hired a voice-impersonator of Midler and carried on with using the song for the commercial, since it had been approved by the copyright-holder. Midler's image and likeness were not used in the commercial but many claimed the voice used sounded impeccably like Midler's."
As a casual mostly observer of AI, even I was aware of this precedent
What was the result of that? Did Ford or Midler end up winning?
Midler won, it’s a cornerstone case in protecting image/likeness.
In tech we’re used to IP law. In entertainment, there is unsurprisingly a whole area of case law on image and likeness.
Tech will need to understand this—and the areas of domain specific case law in many, many other fields—if AI is really to be adopted by the entire world.
I don’t think that’s quite right.
OpenAI first demoed and launched the “Sky” voice in November last year. The new demo doesn’t appear to have a new voice.
I doubt it would take them long to prepare a new voice, and who’s to say they wouldn’t delay the announcements for a ScarJo voice?
A charitable interpretation of the “her” tweet would be a comparison to the conversational and AI capabilities of the product, not the voice specifically, but it’s certainly not a good look.
I believe that "Sky" voice was first released in September last year and according to the blog post released by OpenAI they were working with "Sky" voice actress months before even contacting Scarlett Johansson for the first time.
Does it really cost a to train one voice?
Seems pretty reckless to not have alternatives just in case Scarlett refused.
Probably 5-10 minutes worth of dataset and GPU time for finetuning on an existing base model. Could be done on a Blu-ray rip or an in-person audition recording, legality and ethics aside.
The tweet is so fucking brazen lol
yeah, that was just poking the hornets nest. Even if i wasn’t mad enough to make a stink over my voice before, plausible deniability and all, that would’ve sealed the deal for me.
They approached Johansson and she said no. They found another voice actor who sounds slightly similar and paid her instead.
The movie industry does this all the time.
Johansson is probably suing them so they're forced to remove the Sky voice while the lawsuit is happening.
I'm not a fan of Sam Altman or OpenAI but they didn't do anything wrong here.
Then they should credit that actress and we can see if its legit, otherwise we believe they used copyrighted audio from S. Johansson's movies.
Still live for me? Unless the Sky I’m getting is a different one?
It is. They didn't remove the UI option, they just swapped it out under the hood for the "juniper" voice.
What I don’t understand is what they expected to happen?
Apparently they had no confidence in defending themselves, so why even release with the voice in the first place?
They underestimated how quickly people would take off the headphones and jump on the bandwagon to claim affinity with an injured celebrity.
Are you suggesting they should have engineered the voice actress' voice to be more distinct from another actress they were considering for the part? Or just not gone near it with a 10ft pole? because if the latter the studios can just release a new Her and Him movie with different voices in different geo regions and prevent anyone from having any kind of familiar engaging voice bot.
Smug Silicon Valley entrepreneur. Sam is a trash human
It's not possible for me to express the full measure of my disdain for Sam Altman without violating the HN guidelines.
I still have Sky voice. Is it because of my region?
Well, if any one had doubts Altman is the classic mold of fuck-it-we-know-better-than-anyone tech bro, this is your proof.
Please tell us about the time you most successfully hacked some (non-computer) system to your advantage.
People hire celebrity voice impersonators all the time. You've heard a few impersonators this month probably from ads. This is such a non-issue that's only blowing up because Johannsen wrote a letter complaining about it and because people love "big tech is evil" stories.
I still have the sky voice in my app.