Elon Musk testifies that xAI trained Grok on OpenAI models
TechCrunch • • general
Key Points:
- Elon Musk admitted in a California federal court that xAI partly used distillation techniques on OpenAI models to train its AI, acknowledging this as a common practice among AI companies.
- Distillation involves training new AI models by prompting publicly accessible chatbots and APIs, allowing competitors to create capable models at lower costs, which threatens AI giants' investments in compute infrastructure.
- Musk is suing OpenAI and its executives for allegedly breaching the original nonprofit mission by shifting to a for-profit structure, with the trial currently underway.
- OpenAI, Anthropic, and Google have formed the Frontier Model Forum to share strategies to combat distillation attempts, particularly from Chinese firms, by monitoring and preventing suspicious mass queries.
- Musk ranked Anthropic as the leading AI provider, followed by OpenAI, Google, and Chinese open-source models, while describing xAI as a smaller company with only a few hundred employees.
