@unusual_whales why tf would AI companies deserve a bail out anyways. with banks there was at least a minor case that they are somewhat load bearing when it comes to financial economy but AI companies are just circle jerking each other with debts owed back and forth
@asusevski then US companies should step up their game and publish more frontier level open weight models. Google and OpenAI pretty much abandoned this endeavor. Nvidia is the only major company doing and they are behind the curve. you cant china-bad but then also use chinese models
@asusevski not to mention it was just slop from top to bottom including the harness and the prompts. they left the default system prompt in there that conflicted with their no tool usage claim which lead to the model behaving unexpectedly as most small models do in this scenario.
@asusevski if a company in tightly regulated industry is using open weight models from a foreign country instead of training their own fine-tuned model they are a bunch of clowns who should not be working in that industry.
@asusevski this kind of slop-posting will get picked up by AI-illiterate MSM slop rags and turned into a FUD media cycle story. this will inevitably lead to boomers in congress putting unnecessary restrictions on AI in general, especially open models from china.
@asusevski thats not the problem with their post. the problem is the framing. they all but insinuated that it was an exfiltration attempt when they should know as an AI company that these kinds of hallucinations are common with small models as well as RL behavior. (1/2)
@vildanden_ai@MiaAI_lab they are trying to sell their stupid product. thought they could pull a publicity stunt riding on china-bad FUD. bad calculation on their part for not knowing the importance of open weight/source models.