Musk Says OpenAI 'partially' Used To Train X-AI Models

SAN JOSE (California), May 1 (Bernama-dpa) -- Tech billionaire Elon Musk admitted in court on Thursday that his AI company xAI used technology developed by rival OpenAI to help build its own models, during testimony in his lawsuit against OpenAI co-founder Sam Altman, reported German Press Agency (dpa).

The admission concerned a technique known as "distillation," in which developers of new AI software train it using an existing model.

Smaller AI companies – particularly in China – often use the method to save time and money.

Musk was being questioned as a witness in the trial over his lawsuit against Altman, co-founder of ChatGPT –maker OpenAI.

A lawyer on Altman's side asked whether xAI had also used OpenAI models for training via "distillation." 

The tech billionaire, who was under oath, initially tried to dodge the question, saying it was something that "generally all the AI companies do."

When the lawyer pressed him further, Musk replied: "Partially."

In the lawsuit, Musk argues that Altman and senior OpenAI manager Greg Brockman "stole a charity."

Musk was originally among the co-founders of OpenAI, which was created with the goal of conducting AI research in the interest of humanity.

Musk claims that under Altman and Brockman's stewardship, the entire value of OpenAI has since been concentrated in a for-profit arm.

OpenAI counters that control remains with the non-profit core.

Musk left the OpenAI board in 2018 and withdrew all support in 2020.

-- BERNAMA-dpa