vsreekanti
Throw more AI at your problems
Deep AI Work
Customers want more AI, not better AI
There's no silver bullet in AI
Llama 3.1 405B, Meta's AI strategy, and the new, open frontier model ecosystem
The bull case for AI startups
AI is breadth-first, not depth-first
Open LLMs don't need to beat OpenAI
AI must be more than a checkbox
How to talk to someone who doesn't trust AI
Lessons from customer evaluations of an AI product
The end of the "best open LLM"
You can't build a moat with AI
Untangling concerns about consolidation in AI
Don't build your applications on a single LLM call
Building on quicksand – the state of modern AI app development
The koan of an open-source LLM
LLMs are the variable interest credit card of tech debt
An Introduction to Evaluating LLMs
Prospecting in the AI Gold Rush
Multimodal LM roundup: Unified IO 2, I/O, Gemini, LLaVA-RLHF, and RLHF questions
OpenAI will solve its privacy problems before you
Building a business in Generative AI [video]
Why open-source LLMs matter
A deep dive on MemGPT – memory management for LLMs
Re:OpenAI is too cheap to beat
on Thursday October 12, 2023 @03:25PM (#37861887)
I completely agree — open-source models and custom deployments just can't compete with the cost and efficiency here. The only exception here is if open-source models can get way smaller and faster than they are now while maintaining existing quality. That will make private deployments and custom fine-tuning way more likely.
Re:OpenAI is too cheap to beat
on Thursday October 12, 2023 @02:54PM (#37861422)
Probably some combination of all the above! I think 1 and 2 are interlinked though — the cheaper they can be, the more they build that moat. They might be eating the cost on these APIs too, but unlike the Uber/Lyft war, it'll be way stickier.
Re:OpenAI is too cheap to beat
on Thursday October 12, 2023 @02:53PM (#37861405)
Yep, we agree that the obvious direction of innovation for OSS models is smaller and cheaper, likely at roughly the same quality: https://generatingconversation.substack.com/p/open-source-ll...