💾 Archived View for bbs.geminispace.org › s › AI › 19415 captured on 2024-12-17 at 15:05:07. Gemini links have been rewritten to link to archived content
-=-=-=-=-=-=-
Growth in the LLMs is very rapid. Training of such large models is only possible for the big corporations. The amount of energy required for such training is so huge that it can be equivalent to electricity consumption of a major city for a month. In order to advance these models further, even larger infrasture will be used. Is there any way of reducing this energy requirement ? May be using existing LLMs to build the next generation LLMs ? In any case, what will be the capabilities of next generation LLMs?
Sep 05 · 3 months ago · 👍 50lo · 🙁 1
I think it’s hard to make any predictions in this area. Many research groups work on different directions of LLMs evolution. There is no way to know which one will be successful.
As LLMs start ingesting the garbage they pump out (and in large volumes - even a simple question results in pages of bullshit from these things!)... obviously, the quality will suffer and eventually we will get randomish words strung together.
This can already be observed.
I actually rather enjoy ChatGPT, and use it quite a lot. It is a decent assistant, but you must be an expert in the field in order to detect hallucinations and work around them....
In most new technology, there's always seems to be two new phase. There's the new technology and then often there's a second technology that totally outperformed this first new technology. It feels like this generation of LLMs or AI is only just a proof of concept and we will see very soon in the next weeks or months something that will outperform that technology.
👻 mediocregopher [...] · Sep 06 at 08:42:
I have a much more pessimistic attitude... I think the utility of AI is going to plateau very quickly, if it hasn't already. It can do some things well, other things poorly. The market for the things it does well will get so saturated that it won't be economically viable to run the models which were used to saturate the market. There will be a crash in AI companies at that point. Some will survive and rise from the ashes of that crash and the market will find some equilibrium, but its growth won't outperform the overall growth of the rest of the internet.
I've a feeling that I'm too old for this new tech. I hope it was all hype and will only be a better speech recognition tool in future and nothing more. Let all those AI/LLM corporations crash'n'burn!
But no! There is a use for AI. It's choosing targets in Palestine, so no human is really responsible!