The guy running the leading AI company was previously fired by the board because they didn't think he was showing proper restraint with regard to A LOT of areas. That decision was quickly overturned. And now we have OpenAI showing that they have no qualms about deep-faking a voice actor if they don't get the proper rights.
And we're supposed to believe that OpenAI has the proper protections in place to safely bring into fruition the AI of tomorrow?
The Board of Directors firing him would have had a better chance of sticking if they had been open and clear from the beginning instead of playing some weird Cloak and Dagger game that made them look suspicious. It's quite possible that both Sam and the Board are in over their heads.
Yep agreed, the secrecy is making this all a whole lot worse. I'm looking for to the point when we find out what really happened, assuming that comes out in my life time.
When you handle a sociopath who was already trying to dismantle the board, no, you don’t deal with him in the open by broadcasting your every move. You also try to handle it internally.
We will wish in the future they had succeeded. He’s on the Elon fast track.
> And now we have OpenAI showing that they have no qualms about deep-faking a voice actor if they don't get the proper rights.
No, that isn't established. They are claiming they used a different actress, so no deep-faking involved. If the actress turns out to not be real, then yes Sam A is guilty of lying and some pretty gross behavior. If it turns out to be true though, what they did is a bit scummy, but not dishonest ~and not illegal~, unless you believe that people who sound similar to Scarlett Johansson don't have the same rights to their voice as Johansson does to hers.
Edit: It may actually be illegal given the intent. There is court precedent on this
<not2b> The answer, based on two different court precedents (Bette Midler, Tom Waits), is that the company can't do that. Companies cannot hire soundalike people to advertise their products after the person with a distinctive voice they really wanted declined. Doesn't matter if they hired a soundalike and used her voice.
I think it says something different about intent, rights, and expectations if someone engineers a voice model based on someone's copyright work, versus just finding another real human who's voice is close enough to what you want.
Open AI is already dealing with Microsoft, Microsoft already had a working relationship with Jen Taylor (Cortana), why not contract her and dodge the infringement and creepiness of stealing from Her?
> what they did is a bit scummy, but not dishonest and not illegal
Imitating someones voice doesn't break copyright but there have been several cases against companies that have used sound alikes in (for example) commercials
If Altmans use of "Her" in a tweet and the previous communication between OpenAI and Scarlett Johansson could persuade a judge that it wasn't a random chance they chose an actor that sounded like her but had set out to imitate her voice deliberately she might have a strong case for damages
For example Tom Waits was awarded $2.6M when Frito-Lay used an actor that sounded like him in an ad:
> If it turns out to be true though, what they did is a bit scummy, but not dishonest ~and not illegal~, unless you believe that people who sound similar to Scarlett Johansson don't have the same rights to their voice as Johansson does to hers.
I am so sick of reading this shallow dismissal.
It’s a classic bad faith argument where the offender tries to undermine the victims rights by claiming: “in fact I’m really just defending this other persons rights!”
It’s an intentionally obtuse strawman to say “people who sound similar to”.
It’s more than just Johansson’s voice. It’s meant to evoke the character she played in a film.
A performance that she developed and provided for a role for a project she worked on.
You may be unable or unwilling to appreciate that as work, or even hard work, but it is, and this skeezy company knows it too because they were so taken by that performance that they wanted to use it in their product.
I’ll frame it in a way people on this site can understand.
If you steal my code, but have some third party copy it while changing function and variable names(unfortunately exactly what these llms are also doing) and I complain about you doing that, and you say “O, so I guess anyone who writes python doesn’t have the same rights to their own work as you now?!”
I was unaware that you had insider knowledge of Johansson's contract. Where can I read the sources that supports this assertion?
I suppose in the absence of Johansson's rights to the character she developed means that this unethical megacorp has defacto rights to that character? Absurd.
> It’s a classic bad faith argument where the offender tries to undermine the victims rights by claiming: “in fact I’m really just defending this other persons rights!”
> It’s an intentionally obtuse strawman to say “people who sound similar to”.
> I’ll frame it in a way people on this site can understand.
Perhaps you wouldn't be so sick of it if you attempted to actually understand the argument. The irony in dismissing this as a "shallow dismissal" is you're actually making a shallow dismissal yourself. Your condescending attitude and feeling that you're so much smarter than everybody else is also (I suspect) a hindrance to actually trying to understand what someone with a different viewpoint than yours might possibly be thinking.
This particular case is pretty clear cut: Sam A liked Johansson's voice and either found an actress to mimic it, or he stole it. But when you are creating policy and law, you have to get at the underlying principles, and be able to apply that equally to multiple situations, many of which will be far less clear cut than this one. I.e. laws don't (or shouldn't) say "Sam A can't imitate Scarlet Johannson's likeness," they have to provide some sort of standard and/or benchmark that defines the underlying principle.
Is it possible that you are talking about a specific case, while the people you accuse of shallow dismissals are actually going a lot deeper than you and talking about the greater (and much harder to nail down) principles?
This particular case is boring. It's mostly just gossip mixed with cult of personality. Far, far more interesting to me is the philosophy and principles behind it.
Welcome to HN by the way! (and I mean that seriously. We've obviously disagreed here, but that's what makes this site great IMHO. I do hope that you come to have a little more humility and respect for others though, otherwise you're likely to miss the real interesting deep discussions I love this site for).
> Is it possible that you are talking about a specific case, while the people you accuse of shallow dismissals are actually going a lot deeper than you and talking about the greater (and much harder to nail down) principles?
Bootlicking for an unethical megacorp is the road to "deeper and greater principles"? Peak HN.
This trope is very near the top of my list of things that I wish we could get rid of in conversation. It's a kind of reversal that I suppose is born of frustration rather than being a genuine holistic opinion, but using words like "naive" creates the impression that you would rather criticize people for wanting better behavior from executives than criticize executives for misbehaving. That's a pretty hideous position, and I don't think it's the position you're actually trying to take.
The article is not implying he was ever truly honest. It just saying we have more evidence of his “inconsistent candor” mentioned in his firing letter.
There ain't no naivete, here on HN, for "fault" to be applicable.
If you think ill-behaved billionaires are excusable, then we disagree. Though we seem to be stuck in a dystopian world run by them - so I guess you "win", in a sense.
While I somewhat doubt it was done by design, this is getting far more attention than the GPT-4o announcement did and in the end will likely benefit OpenAI even if they have to make some kind of settlement for misuse of Scarlett Johansson's voice. Wouldn't surprise me though if they knew there would be attention getting blowback and went this way after seeing the GPT-4o announcement mostly get "that's kind of neat" and shrugs.
Attention isn't necessarily always a good thing. Negative attention can cause people to decide to actively avoid that company and switch to a competitor. Just ask Firestone.
It's already easy to put to rest
- They asked her twice
- They tweeted many references to Her
- She declined
- Everyone on Earth heard the voice and thought it was Her
- OAI doesn't give a fuck even tho they could have just picked any other voice that could possibly exist, because they should be able to do anything they please without consideration of humans
I thought the amount of "Weirdo defending Musk on Twitter' energy for Elon was amazing, but the amount of reality-denial that the OAI defenders have is quite a different beast. They wanted SJ's identity, she declined, they effectively cloned her against her will instead. That's about as sick as you can get. Future really looks bright!
No it's not put to rest, because if they used a different voice actress for the model then it's a legitimate move.
Agreed that it's obvious that (assuming there is another voice actress) they found one as close to Scarlett Johansson as possible. That's a dirt bag move. But ~it's not illegal and~ it's not copying if they used a real, different actress. It's Scarlett Johansson's right to refuse her own voice, but it's not her right to prevent anyone else from agreeing.
That voice actress that they used has just as much right to her voice as Scarlett Johansson does to hers. Why should the voice actress not be allowed to work just because she sounds like Johansson? Why shouldn't Johansson be disallowed to work because she sounds like the unnamed voice actress?
Edit: It may indeed be illegal. Thank you child comments
> No it's not put to rest, because if they used a different voice actress for the model then it's a legitimate move.
That is simply untrue. Check out the Midler-Ford case, the Tom Waits case as just two examples.
If the intent is to mimic the character, you're in trouble. OpenAI is trying to claim they weren't mimicking Samatha, but you'd have to be a fool to believe that, given the other evidence.
Very interesting, thank you! Given that I agree it's likely not legal. There is definitely plenty of evidence that they wanted Johansson's voice.
I'm quite torn on whether this is how it ought to be though. If the burden of proof is on showing intent, then I think it's a good balance. If it shifts to just "chance of confusion" like many trademark cases have, that seems pretty unfair to average people and stacked in favor of the rich and famous.
Experience tells me that everyone loves when someone chooses what the law allows over morality.
Can we acknowledge that immoral acts are often not illegal and illegal acts are often not immoral and then say that it's put to rest that it's obviously and definitely a dirt bag move?
> Can we acknowledge that immoral acts are often not illegal and illegal acts are often not immoral and then say that it's put to rest that it's obviously and definitely a dirt bag move?
Absolutely! No disagreement from me there.
I would also add that (to me personally) companies choosing what the law allows over what is moral doesn't impress me at all, in fact it lowers (sometimes drastically) my opinion of the company. When companies choose the moral thing over not-illegal-but-gross, especially when it costs them money, it raises them a huge deal in my book. I try to celebrate those decisions when I see them.
Right, and it was just a slap on the wrist for them. I'm betting OpenAI's bet was that they get the buzz and can take the hit from the lawsuit without breaking a sweat. That is, the expected cost of the marketing stunt factored in the cost of the resulting lawsuits and expected settlement and was signed-off on as acceptable.
I really like the way rusty foster described this in today in tabs, it shows the correct amount of contempt for the exercise itself and derision for how stupid they're expecting people to be to believe it.
> The Torment Nexus announces it has retained counsel in OpenAI dispute.
> On Sunday OpenAI announced it was removing its new horny robot girlfriend voice, explaining in a blog post that the voice was “not an imitation of Scarlett Johansson” at all (and frankly it’s weird of you even to think that???) but was actually the real human meat voice of a completely different actress who goes to another school, you wouldn’t know her. Also she’s from Canada.
> Recapping the imbroglio, Casey Newton wrote that Sam Altman is full of shit [measured, respectful] and Ed Zitron wrote that Sam Altman is full of shit [angry diatribe]. But with the final departure of co-founder and actual technologist Ilya Sutskever, whose last-ditch effort to oust Altman failed last fall, OpenAI is now firmly under the control of Paul Graham / Marc Andreessen style technofascist incompetents, and this kind of clown show will increasingly be the norm.
There's about a dozen links in there I don't feel like copying but it's easy to find.
Does Gizmodos sketchy past discredit the evident discrepancy between Sam's words and deeds? Or do you mean that one excuses the other? I really don't see your point.
Isn't what you did here a perfect example of how – if we were to take it seriously – any discussion could be "stopped" merely by whatabouting about the messenger?
The guy running the leading AI company was previously fired by the board because they didn't think he was showing proper restraint with regard to A LOT of areas. That decision was quickly overturned. And now we have OpenAI showing that they have no qualms about deep-faking a voice actor if they don't get the proper rights.
And we're supposed to believe that OpenAI has the proper protections in place to safely bring into fruition the AI of tomorrow?