news.volyx.in

I genuinely don't understand why some people are still bullish about LLMs (twitter.com)

718 points by ksec · 493 days ago · 1223 comments on HN

Article summary

The author expresses skepticism about the usefulness of Large Language Models (LLMs) due to their tendency to fabricate links, references, and quotes, and their inability to provide accurate sources. Despite using various LLMs, the author has found them to be inconsistent and unreliable. The author questions the value of LLMs and their potential to replace human researchers. The author also mentions that some people believe knowledge graphs will solve the issues with LLMs, but the author disagrees.

Main themes

  • LLM limitations
  • Research accuracy
  • AI hype
  • Language model applications
  • Knowledge graphs
  • Human-AI collaboration

What commenters say

  • Some users have had positive experiences with LLMs, finding them to be useful for specific tasks such as market research and topic deep-dives.
  • Others argue that LLMs are being used incorrectly, and that their limitations are due to user error rather than flaws in the technology.
  • There is a concern that LLMs are being overhyped and that their current capabilities are not living up to expectations.
  • Some commenters believe that LLMs have the potential to be useful, but only for narrow and specific use cases, and that they are not a replacement for human researchers.
  • Others argue that LLMs are already providing significant value, even if they are not perfect, and that their limitations can be mitigated with proper use and training.
  • There is a debate about the potential for LLMs to replace human workers, with some arguing that they will augment human capabilities and others arguing that they will displace certain jobs.
  • Some commenters express frustration with the lack of transparency and accountability in LLM development, and the potential for LLMs to perpetuate biases and inaccuracies.
  • Others argue that LLMs are a powerful tool that can be used to improve productivity and efficiency, but that they require careful evaluation and critical thinking to use effectively.