Google’s AI developer lives rent-free in Elon Musk’s head

About a week in Musk v. Altman Tested, we’ve heard from some of the most powerful people in the technology — including OpenAI President Greg Brockman, Elon Musk’s founder Jared Birchall, and Musk himself. But one of the most prominent is roaming the fringes: Demis Hassabis, CEO of Google DeepMind.

Hassabis is an architect at Google’s in-house AI lab. He founded DeepMind as an independent startup in 2010 and sold it to Google four years later, he says in the middle $400-650 million. Since then, he has been at the forefront of many of Google’s major AI research projects, such as AlphaFold – and climbed the ladder from there, now leading Google Gemini, the team formerly known as Google Brain, and making a profit for the DeepMind spinoff. Isomorphic Labs.

From the beginning, OpenAI was designed to challenge Google. Musk testified that he was inspired to find this in a conversation with Google’s Larry Page, where Page – in his own words – he complained about the matter AI destroys humans. It’s no surprise that Musk and Altman’s team would be wary of its AI team. But court documents and testimony show just how much Google and Hassabis struck fear into their hearts.

During Brockman’s testimony this week, he said Musk talked about Hassabis “frequently” in the early years of OpenAI, calling Musk “very focused and focused” on the man. After attending an AI dinner with Altman and Musk, Brockman added, the first thing he remembered Musk asking was, “Is Demis Hassabis evil?”

The dinner with Demis before OpenAI was “very scary,” Musk wrote in one email to Brockman and Ilya Sutskever, OpenAI’s co-founder. “I feel like they’re playing the Super Bowl and we’re playing the Puppy Bowl. Unless we want our ass handed to us, we need to step up our game dramatically,” he said in 2016.

Hassabis first appeared in the OpenAI courts shortly after its launch, when Musk talked about the “openness” of the new lab in the press. In January 2016, Musk sent Altman and Sutskever, who had been poached from Google, message Hassabis had sent him. The head of Google disagreed with Musk and his co-founders “after praising the benefits of open AI.” Hassabis wrote that it was “very dangerous,” adding, “I think you realize that this is not an alternative solution to the problem of AI in some way?” A few months later, OpenAI co-founder (and current company president) Greg Brockman said Musk that Google’s “legal people” want to talk to him, fearing that OpenAI “will make a public case that it is wrong to have closed AI.” Musk was very interested in who, in particular, called Brockman from Google.

This was the beginning of the competitive years, and things only went up from there.

About six months later, Musk began to express his concerns about beating Google DeepMind in the AI ​​competition. He he wrote to Neuralink colleagues, “Deepmind is moving too fast. I worry that OpenAI is not on track to catch up. Putting it as a non-profit may, in retrospect, have been the wrong move. The sense of urgency is not high.”

In September 2017, Brockman and Sutskever he wrote for Musk is showing his concern about managing OpenAI, using Google as an example of what not to do. “You’re worried that Demis could bring about an AGI dictatorship. We’re also worried about creating a system where you can be a dictator if you choose.”

As of early 2018, Musk appears to have a strong hold on Google AI and the need for OpenAI to defeat the tech giant — and the relative fear has spread to others as well. Musk wrote in January email exchange that OpenAl was “on the path to another Google failure. Clearly something needs to be done quickly or dramatically or everyone but Google will be given unnecessary work.” He and OpenAI co-founder Andrej Karpathy were so concerned that they decided to fold OpenAI into Tesla for better support.

“It’s unclear whether a company can ‘reach’ Google’s scale” without a merger, Karpathy later wrote: “I don’t see anything else that can reach Google’s headquarters in a decade.”

Shivon Zilis, a board member of OpenAI at the time, offered a specific idea. Zilis, who now shares four children with Musk, he wrote a personal request asking him to “reduce” Hassabis. “There is a very low chance of a good future if someone doesn’t slow down Demis. Slowing him down is the only invincible thing I can see,” Zilis wrote. “I think you know that I am not a malicious person, but at the moment I find it impossible to find a way to slow down or change his course.” Musk replied that he would be able to talk on the phone that night but for the first time appeared to be downcast about the prospect of his battle with Hassabis, writing, “I doubt I will be able to do so in a meaningful way.”

Zilis continued to plead with Musk to hire Hassabis, sending rumours from Altman and others. “On top of people secretly chatting on Twitter DM because they don’t trust Demis not to scratch their emails and gchat, the inner circle also meets in a coffee shop in London without mobile phones to discuss privately away from him,” he wrote.

By November of that year, Musk he wrote in an email he “lost confidence” that OpenAI could “act as an opponent” to beat Hassabis and DeepMind, and that he plans to do so through Tesla instead. “We have billions of dollars a year to build tools that hopefully have a chance to keep Google honest,” he wrote. “My evaluation of OpenAl to be compatible with DeepMind/Google without significant changes in processes and resources is 0%,” Musk. he wrote a few weeks later. “Unfortunately, the fate of humanity rests in the hands of Demis…and he’s doing more than that.”

Three months later, in March 2019, the last mention of Hassabis in the show of genius so far comes from the surprise. message sent by Altman to Musk without further details.

“Have a few Demis updates to share,” Altman wrote. Musk agreed to a phone conversation.