Google’s AI engineer lived rent-free in Elon Musk’s head


About a week in Musk v. Altman During the trial, we heard from some of the most powerful people in tech — including OpenAI chief Greg Brockman, Elon Musk’s assistant Jared Birchall, and Musk himself. But one of the most prominent figures hovers around the margins: Demis Hassabis, CEO of Google DeepMind.

Hassabis is an engineer at Google’s in-house AI lab. He founded DeepMind as an independent startup in 2010 and sold it to Google four years later for a reported sum. between 400-650 million dollars. Since then, he has been at the helm of many of Google’s biggest AI research breakthroughs, e.g Alpha fold — and he’s climbed the ladder from there, now leading Google’s Gemini team, the team formerly known as Google Brain, and even the for-profit company DeepMind. Identical laboratories.

From the beginning, OpenAI was designed to oppose Google. Musk testified that he was inspired to find it by a conversation with Google’s Larry Page, in which Page said – in his narration – I ignored this idea Artificial intelligence that erases humanity. Not surprisingly, Musk and Altman’s circle is wary of its own AI team. But court documents and testimony reveal the extent to which Google and Hassabis specifically struck fear into their hearts.

During Brockman’s testimony this week, he said Musk talked about Hasabi “many times” throughout the early years of OpenAI, describing Musk as “very consistent and fixated” on the man. Brockman added that when he attended an AI-focused dinner with Altman and Musk, the first thing he remembers Musk asking was: “Is Demis Hassabis evil?”

Greg Brockman said on stage that Musk was focused on Hassabis

Dinner with Demis before founding OpenAI was “very disturbing,” Musk wrote in one email to Brockman and Ilya Sutskever, co-founder of OpenAI. “I feel like they’re playing the Super Bowl and we’re playing the Puppy Bowl,” he said in 2016. “Unless we want to give up, we need to step up our game dramatically.”

Hassabis first appeared in court documents shortly after OpenAI was founded, as Musk was speaking in the press about the “open” nature of the new lab. In January 2016, Musk sent a message to Altman and Sutskever, who had been stolen from Google, message Al-Hasasis had sent him. The Google chief disagreed with Musk and his co-founders in “extolling the virtues of open source AI.” It’s “actually very serious,” Hassabis wrote, adding: “I assume you realize that this is not some sort of panacea that will magically solve the AI ​​problem?” A few months later, OpenAI co-founder (and current company president) Greg Brockman Musk said That “policy officials” at Google wanted to talk to him, fearing that OpenAI would “build a public narrative that it’s wrong to have any closed-source AI.” Musk was particularly interested in who specifically was calling Brockman from Google.

It was the beginning of years of competition, and the stakes would only get higher from there.

About six months later, Musk began expressing concerns about beating Google DeepMind in the AI ​​race. he books “Deepmind is moving very quickly,” he told his colleagues at Neuralink. “I worry that OpenAI is not on track to catch up. In hindsight, setting it up as a non-profit may have been a wrong move. The sense of urgency is not high.”

In September 2017, Brockman and Sutskver books To Musk, who expressed concern about his control of OpenAI, using Google as an example of exactly what not to do. “You’re worried that Demis could create an AGI dictatorship. So do we. So, it’s a bad idea to create a structure where you can become a dictator if you so choose.”

“I feel like they’re playing the Super Bowl and we’re playing the Puppy Bowl.”

By the beginning of 2018, Musk was seemingly in full circle about the impact of Google’s AI and the need for OpenAI to take on the tech giant — and the relative panic had spread to others as well. Musk wrote in January Email exchange That OpenAl was “on a certain path to failure for Google. It is clear that immediate and dramatic action is needed or everyone but Google will become irrelevant.” He and OpenAI founder Andrei Karpathy were so concerned that they proposed merging OpenAI into Tesla so it would be better resourced.

“It is unclear whether the company can ‘catch up’ to Google’s scale” without a merger, Karpathy later wrote: “I can’t see anything else having the potential to reach sustainable capital on Google’s scale within a decade.”

Shivon Zillis, an OpenAI board member at the time, suggested direct intervention. Zelis, who now has four children with Musk, books A personal appeal asking him to “slow down” Hassabis. “There is a very slim possibility of a good future if someone doesn’t slow down Demis. Slowing him down is the only non-negotiable good action I can see,” Zelis wrote. “I think you know I’m not a malicious person, but in this case, it seems fundamentally irresponsible not to find a way to slow down or change its course.” Musk responded that they could discuss the matter that evening on the phone, but for the first time he seemed depressed about his chances in his fight against Hassabis, writing: “I doubt “I can do it in a meaningful way.”

“Slowing it down is the only good non-negotiable measure I can see.”

Zellis continued her personal pleas for Musk to outperform Hassabis, Transferring rumours From Altman et al. “On top of people secretly chatting on Twitter DMs because they don’t trust Demis not to spy on their email and chats, part of the in-group also meets in a London café without mobile phones to have personal discussions away from it,” she wrote.

By November of that year, Musk… books In an email, he said he had completely “lost confidence” that OpenAI could “serve as an effective counterweight” to taking on Hassabis and DeepMind, and that he was planning to do so via Tesla instead. “We have a cash flow of billions of dollars a year to build hardware that we hope will at least have a dark horse chance of keeping Google honest,” he wrote. “My probabilistic rating for OpenAl being relevant to DeepMind/Google without a radical change in implementation and resources is 0%,” Musk said. books After a few weeks. “Unfortunately, the future of humanity is in the hands of Demis… and they are doing much more than this.”

Three months later, in March 2019, the last mention of Hasabi in trial documents to date came from a mysterious source. message Altman sent it to Musk without providing further details.

“Get some moderated Demis updates to share,” Altman wrote. Musk agreed to talk about the matter over the phone.

Follow topics and authors From this story to see more like this in your personalized homepage feed and receive email updates.


Leave a Reply

Your email address will not be published. Required fields are marked *