AI studying AI? The secret of 'knowledge distillation' according to Garry Tan

By Gi Pyeong Lee
AI Tech Garry Tan Artificial Intelligence Model Distillation
Garry Tan speaking to an audience about the future of AI technology and model distillation
AI Summary

Garry Tan, CEO of Y Combinator, argues that US companies should actively adopt 'distillation' technology, which uses frontier AI models to create smaller, more efficient models.

Imagine you are a student struggling with a very difficult math problem. Sitting next to you is the top student in the class. By observing the process, logic, and know-how the top student uses to solve the problem, you figure out how to solve it faster in your own way. Something similar is happening in the world of Artificial Intelligence (AI). Garry Tan, CEO of Y Combinator and one of the most influential figures in Silicon Valley, is raising his voice to argue that US AI labs should actively embrace this approach.

Why is this important?

Behind the chatbots and voice assistants we use every day are massive AI models. These models were born from the vast ocean of knowledge left behind by humans around the world Source: Garry Tan Urges US Open-Weight AI Labs to Distill Frontier Models, Source: Y Combinator’s Garry Tan advocates for open-weight AI labs. There have been voices calling for bans on the practice of AI models learning from other companies’ AI. However, Garry Tan has offered the exact opposite opinion. He suggests there is no need for regulators to step in and block or regulate this Source: Garry Tan Tells Regulators to Leave AI Model Distillation Alone, Source: Y Combinator’s Garry Tan says ‘do nothing’ about distillation. Why? To ensure the US does not lose its lead in the AI ecosystem and to lower reliance on foreign models.

Easy to understand

In technical terms, this is called ‘knowledge distillation.’ Simply put, it is a training method that extracts knowledge from huge, smart ‘frontier AI models’ (the most advanced models with cutting-edge performance) and transfers it to much smaller, lighter models [Source: Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too TechCrunch](https://techcrunch.com/2026/09/11/y-combinators-garry-tan-wants-u-s-open-weight-ai-labs-to-distill-frontier-models-too/).
Let’s use an analogy. A frontier model is like a university professor who has memorized thousands of textbooks. ‘Distillation’ is the task of distilling the essence of that professor’s lecture into highly efficient summary notes. While these summary notes (the small AI models) might not possess the professor’s vast knowledge, they process information much faster and are easier to carry around. Garry Tan argues that US startups should use this method to create faster and more robust American-made open AI alternatives [Source: Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too TechCrunch](https://techcrunch.com/2026/09/11/y-combinators-garry-tan-wants-u-s-open-weight-ai-labs-to-distill-frontier-models-too/), Source: Garry Tan Urges US Open-Weight AI Labs to Distill Frontier Models.

Current situation

The AI market is currently fierce. Foreign labs are already actively utilizing this distillation technology to quickly catch up to models that US companies have spent trillions of won to create Source: Garry Tan Urges US Open-Weight AI Labs to Distill Frontier Models, Source: Garry Tan Wants U.S. AI Labs to Distill Frontier Models. Garry Tan believes that if US companies do not want to fall behind in this competition, it is vital to share and develop this technology domestically to create a healthier AI ecosystem. He envisions a mutually beneficial structure where companies building massive models continue to pioneer new technologies, while smaller labs take that knowledge to create lighter, easier-to-use models Source: Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too · Issue #831 · hanzhad/squelch-news-engine].

What happens next?

If Garry Tan’s proposal is accepted, we are likely to see a much wider variety of American open AI models in the future. This goes beyond simple competition between companies; it could serve as an opportunity to lower AI dependence on specific nations and allow developers worldwide to use safer and more reliable models. We need to watch how AI models will be developed—not just growing in size, but becoming both smarter and lighter—and what strategy the US will use to protect its AI dominance.

MindTickleBytes AI Reporter’s Perspective

Technology is like a river. No matter how high a dam you build, the flow of knowledge eventually finds its way downward and to wider areas. Garry Tan’s proposal shows that it is much wiser to channel that flow in a beneficial direction rather than trying to block it.

References

  1. Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too TechCrunch (https://techcrunch.com/2026/09/11/y-combinators-garry-tan-wants-u-s-open-weight-ai-labs-to-distill-frontier-models-too/)
  2. Garry Tan on Distillation: Surprising Open-Weight AI Risk (https://www.progressiverobot.com/2026/09/12/garry-tan-open-weight-ai-labs-distill-frontier-models/)
  3. Y Combinator’s Garry Tan wants US open-weight AI labs to ‘distill’ frontier models, too · Issue #831 · hanzhad/squelch-news-engine (https://github.com/hanzhad/squelch-news-engine/issues/831)
  4. Y Combinator’s Garry Tan wants U.S. open-weight AI labs to ‘distill’ frontier models, too - Jingletree (https://jingletree.com/y-combinator-s-garry-tan-wants-u-s-open-weight-ai-labs-to-distill-frontier-models-too-267965.html)
  5. Y Combinator’s Garry Tan wants U.S. open-weight AI labs to ‘distill’ frontier models, too …Middle East (https://eng.pressbee.net/show4875832.html)
  6. Y Combinator’s Garry Tan wants US open-weight AI labs to … (https://tech.yahoo.com/ai/articles/y-combinator-garry-tan-wants-205947138.html)
  7. Garry Tan Tells Regulators to Leave AI Model Distillation Alone (https://startupfortune.com/garry-tan-tells-regulators-to-leave-ai-model-distillation-alone/)
  8. Garry Tan Urges US Open-Weight AI Labs to Distill Frontier (https://www.androguider.com/2026/09/garry-tan-urges-us-open-weight-ai-labs.html)
  9. Y Combinator’s Garry Tan advocates for open-weight AI labs (https://cryptobriefing.com/y-combinator-garry-tan-open-weight-ai/)
  10. Garry Tan Wants U.S. AI Labs to Distill Frontier Models (https://techywired.com/business/garry-tan-wants-u-s-ai-labs-to-distill-frontier-models/)
  11. Y Combinator’s Garry Tan says ‘do nothing’ about distillation (https://www.cnbc.com/2026/09/11/y-combinator-garry-tan-says-do-nothing-about-distillation.html)
AD
Test Your Understanding
Q1. What conclusion did Garry Tan reach regarding AI model 'distillation'?
  • Regulations must be strengthened
  • Do nothing
  • Distillation should be completely banned
Garry Tan argued that regulators should 'do nothing' about the issue of AI model distillation.
Q2. What does 'distillation' mean for AI models?
  • Teaching AI human language
  • Using the learning techniques of high-performance AI models to create smaller, more efficient ones
  • Turning off AI power
Distillation is a training technique that utilizes the learning methods of frontier AI to create smaller models quickly and efficiently.
Q3. Why does Garry Tan encourage US AI labs to use distillation?
  • To reduce reliance on foreign AI models and build more robust American open-model alternatives
  • To make AI slower
  • To significantly increase the cost of AI training
He explained that the goal is to reduce dependence on foreign models and build more robust American open-model options.
    AI studying AI? The secret ...
    0:00