I think people might have the implicit idea that LLM companies will continue to give API access as the models become more powerful, but I was talking to someone earlier this week that made me remember that this is not necessarily the case. If you gain powerful enough models, you may just keep it to yourself and instead spin AI companies with AI employees to make a ton of cash instead of just charging for tokens.
For this reason, even if outside people build the proper brain-like AGI setup with additional components to squeeze out capabilities from LLMs, they may be limited by:
1. open-source models
2. the API of the weaker models from the top companies
3. the best API of the companies that are lagging behind
I think people might have the implicit idea that LLM companies will continue to give API access as the models become more powerful, but I was talking to someone earlier this week that made me remember that this is not necessarily the case. If you gain powerful enough models, you may just keep it to yourself and instead spin AI companies with AI employees to make a ton of cash instead of just charging for tokens.
For this reason, even if outside people build the proper brain-like AGI setup with additional components to squeeze out capabilities from LLMs, they may be limited by:
1. open-source models
2. the API of the weaker models from the top companies
3. the best API of the companies that are lagging behind