xAI launches

Elon Musk announced xAI, a new AI company he would lead, bringing together researchers who had previously worked at DeepMind, OpenAI, Google Research, Microsoft Research and Tesla. The team’s stated prior work included contributions to methods such as the Adam optimiser, Transformer-XL, and batch and layer normalisation, and to systems including AlphaCode, AlphaStar and GPT-3.5 and GPT-4 — credentials assembled to signal that the venture was a serious research effort rather than a side project.

xAI’s announced mission was framed unusually broadly for a commercial AI lab: to “understand the true nature of the universe.” Musk described the company as legally separate from X Corp, the entity formerly known as Twitter, but said it would work closely with X, Tesla and his other companies to pursue that goal — a structure that raised immediate questions, given Musk’s history of criticising OpenAI’s own increasingly complex corporate structure, about where the boundaries between the ventures would sit in practice.

The launch came amid Musk’s public criticism, following his 2018 departure from OpenAI’s board, that the lab he had helped start had drifted from open-source, safety-first principles toward a closed, commercially driven model under Sam Altman. xAI’s launch positioned Musk as a direct competitor to that lab.

The announcement itself carried no product: xAI released no model at launch. It introduced its first model, Grok, that November, four months after the company’s founding was made public, giving the July announcement the character of a talent and intent signal rather than a capability one.