After Calling for a Pause in GPT-4 Research, Musk Buys 10,000 GPUs to Join the AI War

Technology Author: Yunfeng Zhang Apr 18, 2023 12:50 AM (GMT+8)

"I'd rather be optimistic and wrong than pessimistic and right." Musk has never been a techno-pessimist. But since the introduction of ChatGPT, from criticizing OpenAI and Microsoft to signing a joint letter to suspend research, Musk's negative attitude toward AI has led to the misconception that he is on the opposite side of AI.


Just recently, Musk spent tens of millions of dollars for Twitter to buy 10,000 GPUs for one of the company's two data centers, or to move forward with a brand new AI project.

Analysts say there's little need to splurge on data center-grade GPUs if Twitter isn't going to use them for AI. It's just not clear exactly how Twitter will develop AI, which could be based on its own massive data to train language models or to improve its search and advertising business.

Musk's enthusiasm for generative AI is well documented, not only in terms of investing in hardware but also in his desire to bring in more AI talent. In late February, The Information reported that Musk was recruiting for a new research lab to develop a competitor to ChatGPT.

But it must also be admitted that Musk has an unusually cautious approach to AI.

On March 29, a public joint letter sparked an uproar, calling for a moratorium on research and development to training AI more powerful than GPT-4 for at least six months, signed by Musk, Apple co-founder Steve Wozniak and other big names in the tech world.

Musk isn't flagging against AI and taking the opposite side of AI development. After all, his Tesla electric car and recoverable rocket both use AI technology extensively, and he predicted in March that the ratio of humanoid robots to humans "could be greater than one to one," boasting that no one is as good as Tesla at solving real-world AI problems.

However, Musk has retained leeway in his overall stance on AI, placing more emphasis on its social risks.

Outsiders have always questioned his initial intentions, suggesting that he is simply stalling to get the leaders to stop and then get a grip on his own research.

At least, Musk's attitude toward AI in public has been consistent and never optimistic and fearless.

In February, Musk already called for regulation of the AI industry, arguing that there are limits in the technology that should not be crossed, suggesting at that month's World Government Summit that

Cars, airplanes and medicine must comply with regulatory safety standards, but AI does not yet have any regulations to control development. We need to regulate AI safety, which actually poses a greater risk to society than cars, airplanes or pharmaceuticals.

Back in 2018, Musk noted that AI is not necessarily bad, but it can be beyond human control, and that regulating AI is very slow. He also gave the example that it took a decade for seatbelts to become a necessity, despite the obvious number of deaths before then. the timeline for AI will advance much more rapidly.

We can't wait a decade to think that AI is dangerous before we decide to do something about it.

Since ChatGPT came out of nowhere, there has been a succession of entrants. Musk, who cares about Tesla, Space X, super high-speed rail, brain-machine interfaces, and Twitter, won't allow himself to be unconnected with the hottest thing in tech.

And there's always something different about Musk. He may, as he says, genuinely want to usher in an era of safe AI, and as always, he believes he can do it.