In a recent testimony in a California federal court, Elon Musk revealed that his company, xAI, has employed techniques known as "distillation" to train its AI model, Grok, using OpenAI's models. This admission sheds light on a controversial practice where AI firms leverage publicly available chatbots and APIs to enhance their own models.
The ongoing discussion around distillation has primarily focused on its use by Chinese companies, which have been accused of creating competitive AI models at a fraction of the cost of U.S. offerings. Musk's acknowledgment indicates that American firms might also be engaging in similar practices to maintain their competitive edge.
Musk stated that it is a common practice among AI companies to utilize distillation methods, responding affirmatively when asked if xAI had done so with OpenAI's technology. This revelation comes as Musk is involved in a lawsuit against OpenAI, its CEO Sam Altman, and co-founder Greg Brockman, claiming that they have strayed from OpenAI's original nonprofit mission by transitioning to a for-profit model.
The implications of distillation are significant, as it poses a threat to established AI giants by potentially diminishing the advantages gained through substantial investments in computing infrastructure. This allows smaller software developers to create models with comparable capabilities at lower costs. The irony is palpable, considering the ongoing debates about copyright and data usage within the AI industry.
Since its inception in 2023, xAI has sought to learn from leading AI players, particularly OpenAI. While the legality of distillation remains ambiguous, it may conflict with the terms of service set by companies regarding the use of their products.
In response to the challenges posed by distillation, OpenAI, Anthropic, and Google have formed the Frontier Model Forum, aiming to share strategies to combat these practices, particularly those originating from China. Their focus is on preventing systematic querying of models that could reveal sensitive information about their inner workings.
During his testimony, Musk also ranked the top AI providers, placing Anthropic at the forefront, followed by OpenAI, Google, and Chinese open-source models. He characterized xAI as a smaller entity with a workforce of just a few hundred employees.
As the AI landscape continues to evolve, Musk's revelations highlight the competitive dynamics within the industry and raise questions about the future of AI development and collaboration.