4 ms·
What Can LLMs Never Do?
- srbhr 2y agoIt is an interesting article, and a good question, "What can AI never do?" There is an ongoing debate on LLMs and Trust, that we should take a look at as well. How much we can trust LLM and the data to make decisions. :)
- o_nate 2y agoIt's a good article. It seems like the author hedges his bets a little bit on the title question. At the end he allows that someday there could be a combination of LLMs, support scaffolding, and prompt engineering that will be able to tackle even the problems it has difficulty with now. But it was interesting as a tour through some seemingly straightforward problems that even the most advanced current LLMs are bad at, as well as containing interesting speculation on the root cause.