DeepSeek, a Chinese AI research lab, has developed an advanced AI model called DeepSeek-R1 that rivals OpenAI's capabilities in mathematical reasoning and code generation while using fewer resources.

Originally part of High-Flyer (a Chinese quantitative hedge fund), DeepSeek was established in 2023 by Liang Wenfeng to focus on AI development. The company has open-sourced its model and six smaller variants under an MIT license, allowing free commercial use and refinement.

OpenAI and Microsoft are now investigating whether DeepSeek trained their model using OpenAI's model outputs, potentially violating terms of service. David Sacks, dubbed the 'AI Czar,' suggests there is "substantial evidence" that DeepSeek used knowledge distillation - a technique where one model learns from another by asking millions of questions to mimic its reasoning process.

Sam Altman, OpenAI's CEO, expressed concern about copying existing work versus creating new innovations, tweeting: "It is (relatively) easy to copy something that you know works. It is extremely hard to do something new, risky, and difficult when you don't know if it will work."

The situation highlights a complex debate about AI development ethics. While OpenAI defends its practice of scraping internet data for training (including from potentially illegal sources), it argues against knowledge distillation by competitors. An OpenAI spokesperson emphasized their efforts to protect intellectual property and work with the US government to safeguard advanced AI capabilities from competitors.

Related Articles

Previous Articles