Elizabeth Lopatto

At a dinner about AI that Brockman and Altman attended, Musk was late. The first thing Brockman remembers Musk asking is, “Is Demis Hassabis evil?” Hassabis, of course, ran DeepMind.

It was initially supposed to be part of Y Combinator, as a research arm. Looking at a solicitation email from Brockman to then-CEO of Yahoo Marissa Mayer, Brockman writes that donors include Elon Musk, Jessica Livingston, Reid Hoffman, and Peter Thiel. “I’m personally donating $100,000,” he writes. But he didn’t end up donating that. Brockman is speaking very quickly and very softly, and YGR has just scolded him for it.

Brockman is taking the stand next.

I still don’t really understand how this expert helps Musk’s case, if at all.

We are now going through previous statements that Russell has made about Musk — for instance, Russell says that the “age of abundance” Musk talks about with AI robots will be great for Musk because of his pay package with Tesla. Also that Russell would not recommend that Musk be an AI adviser for President Trump.

The expert has testified in front of the US Senate about the dangers of open-sourcing AI systems. “If we open-source AI systems that are unsafe, we dramatically increase the risks,” Russell says. One of Musk’s contentions is that OpenAI is betraying its mission by not open-sourcing its models. Russell is now saying, in response to cross-examination, that open-sourcing can make it easier to remove safety guardrails that have been put in place. “It requires additional and very stringent safety measures.”

I really have no idea what it adds to Musk’s case. It seems to just be a way of running out the clock — but why bother? Is it just a way of giving OpenAI less time to defend themselves?

It’s very boring. Mainly we are establishing that the expert witness provided no specific opinions on OpenAI’s safety.

Because otherwise, this would be the guy to ask about the nonconsensual undressing and MechaHitler. We are getting slowly to the point, I think — which is that Russell suggests safety concerns would slow AI development. “Each company individually feels it needs to be in this race,” he says. “That means they can’t stop and solve the safety problem, which I think some of their employees would like to do, but the overall company police is preventing them.”

His name is Stuart Russell, and he’s getting $4,000/hr for his first 40 hours of working with the team (and $1,500/hr after that). Seems like an expensive way for Musk’s lawyers to run out the clock… Because while we are getting some testimony about AI risks, I don’t really understand how this is relevant to the dispute at hand.