The article clearly disputes this. They hired and worked with the voice actor for Sky months before the first time SJ was contacted, and the voice actor used for Sky never had the movie Her or SJ's name mentioned to her a single time
The Movie Her predates all of this by years, and Sam Altman even tweeted "her"! The OpenAI team are clearly well aware of Scarlett's voice (its inconceivable the majority of the team at OpenAI haven't at least seen part of the film that almost defined their industry). The movie predates all of this by years - of course they knew.
When auditioning actors "months before" they can still look for an actor who guess what? Sounds like SJ, even "before the first time SJ was contacted".
As the actor - I'd likely also be looking to emulate SJ in Her - its clearly what the client was looking for.
> its inconceivable the majority of the team at OpenAI haven't at least seen part of the film that almost defined their industry
Let's not exaggerate. It was a somewhat popular movie, yes, but not really defining and far from the first example of conversational AI speaking woman's voice. There are plenty of examples in movies and TV shows.
If anything, the seminal work in this space is Star Trek casting Majel Barrett-Roddenberry as the voice of computer systems with conversational interfaces, as early as 1987 (or 1986, if she had that role in the Original Series; I don't remember those episodes too well), all the way to ~2008 (or to 2023, if you count post-mortem use of her voice). That is one distinctive voice I'd expect people in OpenAI to be familiar with :).
Also, I can't imagine most people knowing, or caring, who voiced the computer in Her. It's not something that most people care about, especially when they're more interested in the plot itself.
> Let's not exaggerate. It was a somewhat popular movie, yes, but not really defining and far from the first example of conversational AI speaking woman's voice. There are plenty of examples in movies and TV shows.
I'm honestly surprised so many people are making this argument, seemingly with a straight face.
It would have been a pretty weak argument even without the tweet from Altman - it is not exaggeration to say it is the canonical "AI voice companion" cultural artifact in our times, but the opposite, it requires exaggeration to downplay it - but then the CEO's own marketing of the connection weakens the argument past the point of plausibility.
Surely there are better defenses available! But with this line ... phrases like "don't piss on me and tell me it's raining" and "don't believe your lying eyes" keep popping into my mind for some reason ...
> its inconceivable the majority of the team at OpenAI haven't at least seen part of the film that almost defined their industry
rather than
> It is inconsistent that Sam personally wasn't aware
(He obviously was)
I'd agree that Majel Barrett-Roddenberry is the prime example of a computer voice interface for most nerds… but then I looked up when Her was released and feel old now because "surely it's not 11 years old already!"
If "her" wasn't the canonical example of a near-future AI assistant in a film, why then does Sam bother to tweet the single word "her" following launch? I think that film is far more influential than you give it credit for here.
Everyone in tech who saw that tweet knew what it meant - a single word. The tweet doesn't even require additional context or explanation to almost anyone in this industry.
There is also a clear difference in the behaviour of the "computer" in Star Trek vs "her" - what OpenAI shipped is far more like the personality of "her" than the much more straight-laced examples in Star Trek, where the computer was virtually devoid of emotional-sounding responses at all.
What matters is whether OpenAI leadership had the movie Her in mind, and the AI in Her is more similar to LLMs than the Next Generation Star Trek main computer.
There have been cases where it was decided that a person had rights[0] to their distinctive voice, as an extension of the Right of Publicity[1]. For example Midler v. Ford Motor Co.[2], and one or two other cases I've seen mentioned but can't remember.
Midler v. Ford is a poor comparison for this case specifically because of:
1. hiring an impersonator
2. copying a song
3. intending to confuse.
If what OpenAI is saying is true, then none of the conditions apply. I'd say (1) is unlikely, (2) is very unlikely, and (3) is maybe, at least to some degree.
I would suggest that (3) is a solid "yes" given the other communications around this, and honestly the similarity of the final tone of voice.
Very little suggests an intent to confuse more than tweets from company leaders confirming that there was intent to confuse. What is left on the table is whether actual confusion occurred, which is different.
NIL rights are pretty broad, and more like trademark rights than patents or copyrights. The main test isn't similarity, it's whether there is brand confusion. Karpathy and Altman's tweets indicate brand confusion.
Still, this isn't recognized in every state or country, and there aren't many cases yet (although there are laws).
To spell it out in the way that doesn't require a crazy level of suspension of disbelief:
Based on the date (right after the public release of the assistant), it is actually very strong evidence for "we thought it would be awesome to have an AI voice companion that sounds like the one in Her", which, combined with the (undisputed) revelation that they indeed tried (twice) to get the person who did that voice, is a very strong indicator of the intent of the thing that happened many months before.
Yes, actually trying to hire Scarlett Johansson is very good evidence that they thought it would be awesome if they succeeded in hiring her as a celebrity voice. A one-word tweet adds nothing to that.
To find this "extremely hard to believe", you have to argue that this story, which has multiple sources unaffiliated with OpenAI, contemporaneous documentary evidence, and is written by a reporter with every incentive in the world to roast OpenAI, is directly wrong about facts plainly reported in the story.
I think you have to want this story to be wrong to think it's wrong. It's a tough beat! Everyone was super sure OpenAI did this bad thing just a couple days ago, and now they're feeling sheepish about it.
OpenAI reached out to SJ to use her as a voice actor.
Why was the casting call for non-union actors only? SJ is a SAG union member. "Non-union actors, unless you happen to be her?"
Her agent feared for her client's safety? Why? And to the point where the agent wants to stay anonymous, too?
Highly ironic, given that ultimately, a single voice statement from the voice actor could be far more conclusive of OpenAI's defense.
Altman was not "intimately involved"... well, other than the latter attempt to sign SJ two days prior to launch coming from him personally...
The agent said the actress confirmed that neither Johansson nor the movie “Her” were ever mentioned by OpenAI. Weird they'd not say anything to her when they were making these references public.
Out of all journalistic outlets, WaPo, being owned by Bezos, has the least incentive to roast OpenAI in this case.
Wouldn't it be nice if Bezos/Amazon could make Alexa sound like a voice actress that sounds a lot like ScarJo without repercussions? First step is to shape the public opinion.
It isn't the specific facts covered by the article that I find hard to believe. It's claims like "they probably weren't even aware of the movie and definitely weren't trying to create a voice that sounded like it".
It's true that their behavior is less damning, with the facts from this reporting. And that's a good thing. But it's not true, in my opinion, that the article demonstrates that there is no behavior to criticize.
I wrote a comment to this exact effect yesterday. We didn't have both sides of the story, let's give Sam/OAI, like, literally just a single day to present their own side. And now he did, and I am sure all the people shouting down Sam at every turn still give no shits and will move the goalposts.
>The agent, who spoke on the condition of anonymity to assure the safety of her client, said the actress confirmed that neither Johansson nor the movie “Her” were ever mentioned by OpenAI.
So an anonymous statement from a puff piece when multiple verifiable public records show OpenAI publicly mentioning Johansson and Her. Count me skeptical.
