Piero Bosio Social Web Site Personale

@mrjunge If you have to spend almost as much time, or more, cleaning up an LLM-assisted product as you originally saved, I agree. LLM's can certainly generate spam or spam-like output at times. But to claim that that's ALL it does is not correct. One company found that initially, their software code writing gains were eaten up by quality control rework, but that after learning how to better use the tools, when to use them, and when not to use them, production went up 30% with no loss of quality.

@leadegroot @dragonfrog @ThePowerNap
I can certainly see a bubble bursting, but I'm not sure that Phind is a sign. Hard to compete with those with billions to develop core models, and not going to hit any profit home runs reselling access to other's models.

@davidgerard @cstross The article you posted included a very useful study. However, contrary to the article, that study is NOT the only study on the topic. Other, far larger, more comprehensive, studies show measured, meaningful productivity improvements. You asked for measurements, so here they are: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4945566 and https://www.youtube.com/watch?v=tbDDYKRFjhk The Stanford study found that the average productivity boost is significant (~20%) but some teams see productivity decrease.

@ThePowerNap @cstross I don't disagree. I think part of the path going forward will be to push more to the edge, as we've seen with other types of AI, using less powerful, but good enough, models for many applications. Models that don't require as much energy to run.

@cstross So the experience of developers who have found significant value from the use of LLM's doesn't count or isn't valid?

Used to WRITE material, they aren't great. But to organize, summarize, do research and help brainstorm, they can be useful.

While ChatGPT currently can't pass the IRS' volunteer tax assessor's exam (see https://www.mcgurrin.info/robots/8298/), I have confidence that a RAG based system could do so.

Heck, COMPUTERS are magic to most people, as are the internet, radios and TV.

@VictimOfSimony @cstross If you want to debate whether or not, as a whole, LLMs are a net positive or net negative, I agree. Intelligent people can legitimately disagree. One cannot, based on the evidence, legitimately claim that they only produce spam.

@rupert @cstross Agreed. Doesn't make them only spambots either.

@cstross Love your fiction, can't agree on this take. Spam does not help with writing code, whereas LLM's can be extremely helpful.

LLM's are not generally intelligent. They can be immensely problematic, but to dismiss them as merely spam generators as you have, or "parrots" as others have is simply incorrect.