Elon Musk’s only expert witness at the OpenAI trial fears an AGI arms race

When bash we instrumentality AI doomers seriously?

That’s a cardinal subtext of Elon Musk’s effort to unopen down OpenAI’s for-profit AI business. His attorneys reason that the enactment was acceptable up arsenic a foundation focused connected AI safety, and mislaid its mode successful pursuit of lucre. To beryllium that, they mention aged emails and statements from the organization’s founders astir the request for a public-spirited counterweight to Google DeepMind.

Today, they called their lone adept witness: Peter Russell, a University of California, Berkeley machine subject prof who has studied AI for decades. His occupation was to connection inheritance connected AI, and found that this exertion is unsafe capable to interest about.

Russell co-signed an open letter successful March 2023 calling for a six-month intermission successful AI research. In a motion of the contradictions here, Musk besides signed the aforesaid letter, adjacent arsenic helium was launching xAI, his ain for-profit AI lab.

Russell told jurors and Judge Yvonne Gonzalez Rodgers that determination were a assortment of risks associated with the improvement of AI, ranging from cybersecurity threats to problems with misalignment and the winner-take-all quality of processing Artificial General Intelligence (AGI). Ultimately, helium said that determination was a hostility betwixt the pursuit of AGI and safety.

Russell’s larger concerns astir the existential threats of unconstrained AI didn’t get aired successful unfastened tribunal aft objections from OpenAI’s attorneys led the justice to bounds Russell’s testimony. But Russell has agelong been a professional of the arms-race dynamic created by frontier labs astir the globe competing to scope AGI first, and called for governments to modulate the tract much tightly.

OpenAI’s attorneys spent their cross-examination establishing that Russell wasn’t straight evaluating the organization’s firm operation oregon its circumstantial information policies.

But this newsman (as good arsenic the justice and the jurors) volition beryllium weighing however overmuch worth to enactment connected the narration betwixt firm greed and AI information concerns. Virtually each 1 of the OpenAI founders person strenuously warned astir the risks of AI, portion besides emphasizing the benefits, attempting to physique AI arsenic accelerated arsenic imaginable — and hatching plans for AI-focused for-profit enterprises they would control.

From the outside, a wide contented present is the increasing realization wrong OpenAI aft its founding that the enactment simply needed much compute walk if it was to succeed. That wealth could lone travel from for-profit investors. The founding team’s fearfulness of AGI successful the hands of a azygous enactment pushed them to question the superior that yet tore the squad apart, creating the arms contention we cognize today—and bringing america to this lawsuit.

The aforesaid dynamic is already playing retired astatine a nationalist level: Senator Bernie Sanders’ propulsion for a instrumentality imposing a moratorium connected information halfway operation cites AI fears enunciated by Musk, Sam Altman, Geoffrey Hinton and others. Hoden Omar, who works astatine the commercialized enactment the Center for Data Innovation, objected to Sanders citing their fears without their hopes, telling TechCrunch that “it is unclear wherefore the nationalist should discount everything tech billionaires accidental but erstwhile their words tin beryllium recruited to capable gaps successful a precarious argument.”

Now, some sides of the lawsuit are asking the tribunal to bash conscionable that: instrumentality portion of Altman and Musk’s arguments seriously, but discount the parts that are little utile for their ineligible argument.