I never bought into the narrative that OpenAI and Anthropic are in trouble because companies can just use "free" Chinese models.
Two problems:
1. You still need extremely expensive hardware to inference them
2. You have to assume that Chinese LLM labs don't care about revenue.
#2 is an especially troublesome assumption because if OpenAI and Anthropic make way more revenue, they can train even better models faster and serve them to even more users - it's a flywheel effect. Therefore, no one should think that these Chinese LLM labs will be "free" forever.
And there it is. It will never be "free" forever.
I never bought into the narrative that OpenAI and Anthropic are in trouble because companies can just use "free" Chinese models.
Two problems:
1. You still need extremely expensive hardware to inference them
2. You have to assume that Chinese LLM labs don't care about revenue.
#2 is an especially troublesome assumption because if OpenAI and Anthropic make way more revenue, they can train even better models faster and serve them to even more users - it's a flywheel effect. Therefore, no one should think that these Chinese LLM labs will be "free" forever.