OpenAI

Look, I’m not bought in on AGI at all, and the “AGI readiness” team getting disbanded in 2024 happened as it became clearer to everyone but the AI cultists that AGI wasn’t possible. (It was clear to some of us from the jump.) I have no idea how this is landing for the jury, but getting safe, beneficial AGI is silly if AI superintelligence isn’t possible.

She initially worked on the “applied” team, but then moved into a research team because it was “more interesting” to work on the “policy” and “AGI readiness” teams, and think about what to do in the case AGI happened. I also prefer daydreaming to actually working.

They also discussed Dario Amodei becoming CEO of OpenAI. “I thought it was an option worth considering among our set of difficult options,” Toner says.

Neither Altman or Brockman had been allowed to tell their side of the story, nor were their HR files pulled by the board. There was no input from Microsoft, or any other investors or customers. Toner smiles when she’s frustrated or annoyed, which she sometimes is by this line of questioning.

I am expecting a relatively sedate day today. We’re going to see more of former OpenAI board member Helen Toner’s deposition. Right now lawyers are discussing when the case will end; we expect closing arguments a week from today.

It was primarily because Altman was not entirely candid with the board about his interests in an OpenAI startup fund. There was also some drama about Toner’s paper, which Altman told Sutskever that another board member suggested Toner resign from the board. That board member said she’d never said it. Further, Mira Murati and Sutskever also mentioned problems. And, of course, the lack of disclosure of ChatGPT...

She says the starting point was Sutskever reaching out to have a conversation where he expressed serious concerns about Altman. It was a “pattern of behavior” that included issues with “honesty and candor” that led to the firing, not any one action. Toner has already laid out some of this in a 2024 podcast, and it’s similar to Murati’s testimony.

She wasn’t surprised she hadn’t been told, though, because “I was used to the board not being very informed about things.” She says that “caused me to believe that [Altman] was not motivated to help the board perform the oversight role.”

That means there’s no clear-cut way to test for safety. People are just throwing things together to see what happens. She refers to OpenAI’s safety board’s methods as becoming “somewhat less slapdash” over time.

This should be about an hour. YGR has told the jury that if she sees them falling asleep, she’s going stop the video and have them stand and stretch.

Every time a MSFT lawyer gets up to question a witness in Musk v. Altman, it’s “And Microsoft wasn’t there?” with an occasional addition of “And Satya Nadella wasn’t there either?” This gets funnier every time it happens.

She is asked about texting Musk about the Microsoft deal with OpenAI — that the structure was not maximum profit and Microsoft was not in control. She looks at the evidence, and says she sees it there but... “it’s not in my neurons, it’s not in my brain, but I see it.” Okay.

OpenAI says it partnered with AMD, Broadcom, Intel, Microsoft, and NVIDIA on a protocol called Multipath Reliable Connection, or MRC, which “improves GPU networking performance and resilience in large training clusters.” The full spec is available through the Open Compute Project.

Zilis said she now recalled certain messages that she had said she didn’t recall in her deposition, saying that at this point she’d reviewed documents numerous times. Eddy said, “Your long-lost memories have since been recovered.”

Three were Tesla AI. One was OpenAI as a B-corp subsidiary of Tesla. One was Altman as anchor for TeslaAI. But my favorite? “Find a way to get Demis. Seriously…. Demis really does fanboy hard and I don’t think he’s immoral… just amoral. If he hung around E perhaps it would force him to think about humanity more.” Hassabis is really haunting these guys.

as part of his push to increase Tesla’s AI presence. “Those who want to work on large scale AI research don’t currently think of Tesla, and Elon wants to change that by announcing his intention to create a world-class AI lab,” Zilis wrote in 2017.

We are seeing more details about Zilis advocating for Musk’s plan to wrap OpenAI into Tesla. “Tesla solves the funding issue immediately… Tesla at least has option to bury,” reads one email from Zilis. “They haven’t internalized the advantages to burying this in Tesla for stealth advantage,” reads another.

They didn’t want Musk — or anyone — to have control over OpenAI. “You and I can argue that’s stupid all we want but they are holding firm on it,” Zilis says in an email to Jared Birchall in September 2017.

That’s what I’ve learned from his emails. Me too, Sam!

A separate ideation email to Altman lays out certain options for changes to OpenAI’s structure, including one option of rolling all of OpenAI into a B-corp, or a for-profit company with a public mission, and another option of having both an OpenAI C-corp and a nonprofit.

Zilis’ answers on the stand are often slightly different from the ones she gave in her deposition. We’ve now had two videos played in the courtroom. She’s also being represented by Musk’s lawyers.

The message, according to exhibits read aloud in court proceedings, said, “I just wanted to say I hope you are [OK]. I have no idea what’s going on but … I care about you as a person first and foremost. Sending all of my positive vibes your way.”

Altman and Brockman were both investors in the nuclear energy company, and since the company didn’t have an official product yet, she said that OpenAI potentially entering into a deal with Helion “felt super out of left field … How is it the case that we want to place [a] major bet on a speculative technology?”

under better light, Zilis’ top is green and not gray.

First, she says that the broad release of ChatGPT wasn’t discussed with the non-profit OpenAI board. This was discussed in a board meeting. Second, the deal with Helion raised eyebrows because Altman and Brockman both had investments and the tech was still speculative. She also felt that “it was probably the only time where I remember feeling in the pit of my stomach -- just being like, I voiced my concerns.”

She said she and the “entire board had voiced extreme concern about that whole massive thing happening without any semblance of board communication.” That was the first concern she raised internally about Altman, she said.

Shivon Zilis recalled Microsoft CEO Satya Nadella saying of OpenAI at the time that Microsoft was “below them, above them, around them.” Zilis said this denoted complete control, calling it “terrifying because [it] was just not the thing that we had been fighting so hard for.”

She said he goes into “maniac mode” and had trouble thinking of a more quantifiable amount of hours. She added, as an answer to a follow-up question, that their time spent together was a brief break from “the insanity … that is his entire work life.”

She is wearing a black cardigan and black pants with a gray shirt. She is saying that after graduating from Yale, she took a job at IBM, then joined Bloomberg Ventures, and launched Bloomberg Beta, where she focused on AI investments. “I was 13 years old and I read a book called The Age of Spiritual Machines by Ray Kurzweil,” and that opened a new world for her. “I read it 10–15 times.”

Sam Altman wasn’t making decisions quickly enough and dragging things out, or not making decisions on controversial things at all, Murati says. He told people what they wanted to hear. “Those are the big themes.”

Satya Nadella met with the OpenAI board. And Murati told Microsoft’s Kevin Scott about Ilya Sutskever signing a petition to reinstate Altman. Murati also wanted Altman back. “The board had not followed a process that could be trusted and it wasn’t transparent with regard to firing Sam.”

It is specifically about how OpenAI and Microsoft worked together. She says that when GPT models were developed, there was no sense they would be commercializable. She also says that Altman undermined her in her ability to do her job, and pitted executives at OpenAI against each other.

Closing statements are likely next Thursday, not early next week. Also, Musk’s lawyer Steven Molo suggested that OpenAI came up with “small adjunct” as a phrase while cross-examining Brockman. In fact, Musk used that phrase.

YGR has sternly warned the lawyers about how much time they have left. We expect closing arguments a week from tomorrow.

The two companies’ famed 2019 contract was made public as part of the Musk v. Altman trial exhibits. The 36-page agreement defines artificial general intelligence as “a highly autonomous system that outperforms humans at most economically valuable work.”

OpenAI may have pledged to cut back on “side quests,” but making its own phone doesn’t count. Right?

Wu is still identifying documents. Two jurors have some real thousand-yard stares going on.

We are now looking at assorted legal documents with him. We have determined the board has approved the agreement of 2018, which is the initial formation of the for-profit OpenAI LP. OpenAI Inc. contributed assets and in result got the limited partner interest and “residual interest.” This seems to be mostly about reading stuff into the record.