3 ms·
LLMs and Agents Are Overhyped, with Dr. Andriy Burkov
- aurareturn 2y agoI personally think they're underhyped. All the skeptics make the same arguments such as hallucinations. Meanwhile, LLMs and agents are making me more and more productive every day. I think all the skeptics are missing the forest for the trees. LLMs and agents don't have to be perfect to be transformative. They're already super useful and just need to be scaled up with more compute capacity, better UX, and better integrated with proprietary data.
- croes 2y agoIf you consider the costs of training and running these models yout productivity boost falls flat. And many people already use these AIs but there isn't any spike in revenue visible.
- aurareturn 2y ago>If you consider the costs of training and running these models yout productivity boost falls flat. I'd like to see your math. >And many people already use these AIs but there isn't any spike in revenue visible. Existing businesses won't see a spike in revenue. They should see better margins though.
- croes 2y agoHow much profit does you productivity boost generate? Compared to the billions of loss from OpenAI and all the other companies. Plus all the damage from better spam and fishing mails. Plus all the damage from fake video and audio. Do we see a spike in profit?
- aurareturn 2y agoBillions of loss is in capex investment and R&D to drive more value. As far as I know, the gross margins for OpenAI inference is in the high 70s last I remember. Profit and loss is not directly comparable to productivity because profit and loss is highly driven by competition.