Comment by wolfgangbabad

In my opinion the LLMs in the "cloud" reached plateau. Still the same probabilistic hallucinations. No idea what they gobbled up. It's a non-reliable circus type of thing. In fact I more and more go directly to Wikipedia or Grokipedia (if you want to know about Odin etc). There is more and more guardrails and talkbacks. They try to make it more human which makes it even more unusable. My guess, more and more marketing people get involved and things turn to unusable shit. I am more optimistic about local llms in a few years where there won't be any guardrails and you will be able to tweak the model how you like and not some marketing specialist at OpenAI or Anthropic.