作者讨论了美国AI公司使用更便宜的国产模型如DeepSeek，建议美国通过KYC认证和更好的推理模型进行创新，并指出Mythos作为仅限美国的、正被蒸馏的前沿模型。
作者讨论了美国AI公司使用更便宜的国产模型如DeepSeek，建议美国通过KYC认证和更好的推理模型进行创新，并指出Mythos作为仅限美国的、正被蒸馏的前沿模型。
“just throwing out a random thought, the UBS report is anecdotally true + and it's interesting there's more Chinese reports on Anthropic distillation. Many US startups/companies using AI: Are using Chinese models like DeepSeek since they’re much cheaper than Gemini/OpenAI/Anthropic reasoning models for cost per task. Kinda a catch 22 when US is built on capitalism and everyone goes for the cheaper options. I think the US just has to get good and out innovate here: - both with heavy KYC-auth frontier models for US only like Mythos given those are being distilled by competitors. and amybe staggered tiers for allied partners so US is always ahead Maybe something like dedicated AI Google auth like FaceID + Persona would do the trick before access/key creation for the newer models, then short lived scopes. At least addressing those simpler fake/stolen account + pool allegations, requiring biometrics before use. Then US Gov drafting new regulations making it illegal for sharing access similar to identity forgery with bank accounts. - then the most obvious one is just heavily optimized reasoning models that mogs DeepSeek and others. Kinda good Trump admin paused Fable/Mythos access, imo not quite sure why a trillion dollar company keeps letting itself get distilled over and over again. There should be friction added if you’re accessing something close to ASI. TLDR: US just needs to good at cheap inference and add bank-grade authentication to AI model access. Easier said than done tho.”原帖：X / @aleabitoreddit ↗