|
|
|
|
|
|
by bg24
34 days ago
|
|
|
Relevant article - https://www.anthropic.com/news/detecting-and-preventing-dist... (3 labs generated over 16 million exchanges with Claude through approximately 24,000 fraudulent accounts). So extraction in this context is distillation. While it is obvious to many, a modern LLM is built in roughly three stages: the foundation (pretraining) model, then SFT/supervised fine-tuning (distillation makes it easy), then the RL/RLHF stage on top (most effort-intensive). For today's reasoning models, RL/RLHF is becoming the most compute-intensive part. Companies like Anthropic spent millions building those fine-tuning examples. A follower can shortcut that on both cost and time by distilling, and it will keep happening: every time the frontier lab climbs higher, others will find a way to shortcut the new gap. There's very little Anthropic can do beyond fraud prevention and blocking accounts that violate their terms of service. On the policy question, I'm completely against banning Chinese models. I'm a heavy Claude Code user and I'll keep being one. But there should absolutely be price competition. China is eating the rest of the world for breakfast, lunch and dinner on manufacturing, and it did not help to ban them. Frontier pricing can't sit at 10x a capable competitor. It doesn't need to be at par either — demand is higher, and quality, trust, and fewer tokens to finish a task are worth a premium — but 4–5x is defensible. |
|
|
The generalization of this is: technologically advanced societies only continue to function as long as you prevent people from circumventing the technological business model (initial R&D investment that is recouped by selling units of the product above their manufacturing cost) by stealing your R&D (allowing them to sell units based on manufacturing cost alone, because they externalized their R&D to you).
This means both taking action against malicious actors inside your system of governance (IP laws in your country) and outside of it (sanctions, internet blocks, ITAR restrictions).
An honest competitor is perfectly capable of competing on price without stealing IP - see Mistral and that newer EU model that's trained on actually licensed content.
And I agree - we want a wide variety of models, from less-capable (but far cheaper) to those that maximize intelligence at any cost.
But those advocating for China distilling US models are just advocating for wealth transfer from the latter to the former - and highly likely to be 50 cent party members.