(2024-05-28) ZviM OpenAI Fallout

Zvi Mowshowitz: OpenAI: Fallout. Previously: OpenAI: Exodus (contains links at top to earlier episodes), Do Not Mess With Scarlett Johansson... We have learned more since last week. It’s worse than we knew.

The Story So Far

For years, employees who left OpenAI consistently had their vested equity explicitly threatened with confiscation and the lack of ability to sell it, and were given short timelines to sign documents or else. Those documents contained highly aggressive NDA and non disparagement (and non interference) clauses

Here is Altman’s statement from May 18, with its new community note.

Evidence strongly suggests the above post was, shall we say, ‘not consistently candid.’

Ilya Sutskever and Jan Leike, the top two safety researchers at OpenAI, resigned, part of an ongoing pattern of top safety researchers leaving OpenAI.

OpenAI created the Sky voice for GPT-4o

A Note on Documents from OpenAI

Some Good News But There is a Catch

OpenAI has indeed made a large positive step. They say they are releasing former employees from their nondisparagement agreements

These actions by OpenAI are helpful. They are necessary. They are not sufficient.

First, the statement is not legally binding

It also does not mention the ace in the hole, which is the ability to deny access to tender offers, or other potential retaliation by Altman or OpenAI

Nor does it mention the clause of right to repurchase for ‘fair market value’ that OpenAI claims it has the right to do, noting that their official ‘fair market value’ of shares is $0.

Then there is the problem of taking responsibility. OpenAI is at best downplaying what happened. Certain statements sure look like lies

How Blatant Was This Threat?

*Oh my.

Neel Nanda (referencing Hilton’s thread): I can’t believe that OpenAI didn’t offer any payment for signing the non-disparage, just threats…

This makes it even clearer that Altman’s claims of ignorance were lies – he cannot possibly have believed that former employees unanimously signed non-disparagements for free!*

It Sure Looks Like Executives Knew What Was Going On

Kelsey Piper’s Vox article is brutal on this, and brings the receipts. The ultra-restrictive NDA, with its very clear and explicit language of what is going on, is signed by COO Brad Lightcap

Pressure Tactics Continued Through the End of April 2024

The Right to an Attorney

Here we have OpenAI’s lawyer refusing to extend a unilaterally imposed seven day deadline to sign the exit documents, discouraging the ex-employee from consulting with an attorney

The Tender Offer Ace in the Hole

The Old Board Speaks

OpenAI Did Not Honor Its Public Commitments to Superalignment

OpenAI Messed With Scarlett Johansson

As I covered last time, if you do a casting call for 400 voice actors who are between 25 and 45, and pick the one most naturally similar to your target, that is already quite a lot of selection.

Remember, in addition to all the ways we know OpenAI tried to get or evoke Scarlett Johansson, OpenAI had a policy explicitly saying that voices should be checked for similarity against major celebrities, and they have said highly implausible things repeatedly on this subject.

Another OpenAI Employee Leaves

Gretchen Krueger resigned from OpenAI on May 14th

OpenAI Tells Logically Inconsistent Stories

A pattern in such situations is telling different stories to different people. Each of the stories is individually plausible, but they can’t live in the same world.

OpenAl’s messaging has turned into a particularly tangled mess of conflicting promises

When You Put it Like That

Individuals paying attention are often… less kind.

Ravi Parikh: If a company is caught doing multiple stupid & egregious things for very little gain...It probably means the underlying culture that produced these decisions is broken

There is a Better Way

Eliezer Yudkowsky: Departing MIRI employees are forced to sign a disparagement agreement, which allows us to require them to say unflattering things about us up to three times per year. If they don’t, they lose their OpenAI equity

Should You Consider Working For OpenAI?

OpenAI says it should be held to a higher standard, given what it sets out to build. Instead, it fails to meet the standards one would set for a typical Silicon Valley business.

The working conditions? The nature of upper management? The culture you are stepping into? The signs are not good, on any level. You will hold none of the cards.

The Situation is Ongoing

It seems increasingly plausible that it would be in the public interest to ban non-disparagement clauses more generally going forward

This is distinct from non-disclosure (NDA) in general, which is clearly a tool we need to have. But I do think that, at least outside highly unusual circumstances, ‘non-disclosure agreements should not apply to themselves’ is also worth considering.