3 ms·
If you go the GPT route, try to make the backend flexible instead of marrying it to ChatGPT's API. Competing services (if not self hosted LLMs) could drive pric
by brucethemoose2 4y ago
If you go the GPT route, try to make the backend flexible instead of marrying it to ChatGPT's API. Competing services (if not self hosted LLMs) could drive prices down very soon.
Also, those ideas are not mutually exclusive.
- jdbohrman 4y agoYeah that's another thing I'm thinking about right now. Any specific services to look out for?
- brucethemoose2 4y agoFacebook's llama model, which has already leaked out, lol.
- jdbohrman 4y agoOh interesting. I'll check that out!
- brucethemoose2 4y agoAlso to elaborate on what I meant earlier, maybe you could "cache" GPT results and reduce costs by letting users publish the better GPT results on your service as templates.
- jdbohrman 4y agoThat's a SUPER interesting idea!