Schools need to be teaching these kids how to properly use the tools. Without that, they’ll just default to whatever makes their life easiest, which is just having it do everything for them. Those that use it as a tool to aid their thinking, planning, research, etc will come out ahead of the rest

Paper here: https://www.sciencedirect.com/science/article/pii/S294988212500091X

@ValerioCapraro Sort of like this AI-generated post?

@ValerioCapraro Creativity is not a thing. It's just a hail Mary mashup of ideas that may or may not work.

@ValerioCapraro @GaryMarcus This a big point in coding. Code is not just a way to get a feature. Is a way to formulate and comunicate a problem. An with this, is a way to understand it. The way of understanding it gives way to other ways to see the problem, i.e. creativity

@jnobleband @ValerioCapraro We believe that this document is fully AI-generated

https://www.pangram.com/history/37306f9d-3af6-46f6-a784-a033e4adb20d

@ValerioCapraro @pangram slop (ironic slop? Meta-slop?)

@ValerioCapraro Using AI to write admissions essays is a misuse of the tech. Anyone who uses an AI to write their essays is uncreative by definition.

AI tutors, precisely matched to a each individual student's talents and weaknesses would be a better use of the tech.

@DesignCntrl @ValerioCapraro Imagine being this much of a soulless, empty ghoul. Pathetic isn't even a strong enough word here. This is mortifying.

"GPT-4 can write well, sometimes better than individual students."

All LLMs can ALWAYS write better than any human. There's simply no way a human can beat an LLM at writing. In this respect, LLMs are great equalizers especially for those who can't write well (myself included).

They are unbelievably fantastic tools. I am creative so they don't need to be. I am the guide and they my pen. To be frank, I absolutely loathe writing and such a big deal is made out of writing well. Sure, we should all write at an acceptable standard, but writing should not determine who succeeds or doesn't, especially non-native English speakers.

I really doubt your claim for one simple reason. People don't think the same.They don't come from the same background.They don't come from the same lived experience. AI doesn't write for them.They are the guide. It can improve the writing. It can help their thinking, but it doesn't think for them.

Unless they choose to let that happen.

@ValerioCapraro @GaryMarcus Did you use ChatGPT to write this post?

@CalculusNe75865 @ValerioCapraro If you loathe writing, I doubt you've done much thinking of what makes it great.

@Erickschultz11 @ValerioCapraro How does it improve their writing without doing any thinking for them? (Or, rather, doing something for them that would require thinking if they did it themself (because AIs don't actually think at all).)

@DullClayCutter @ValerioCapraro You proved my point.

@ValerioCapraro Only read parts of the methodology and it seems to go in a decent direction. I suspect the research needs to expand to other models (especially older ones and open source ones), worth checking if different models changes semantic differences.

@PeterRex @ValerioCapraro Have you even read the post? There is NO proper way to use those tools. They always converge, no matter the input

@ValerioCapraro https://github.com/Martin123132/This-Parrot-is-no-more-/commit/93d1860f2d64a17119a7ca7b91ee28c1403565a4 The exact same thing has been happening in production for the last 30 years, and people complain that’s there’s a lack of skilled workers 🤣🤣 well it’s not hard to work out why is it, remove the roads, then you have nothing to drive on.

@ValerioCapraro llms and the human hive mind

imo a result of scaled intelligence since ce 1440

printing press -> tv news -> internet -> llms

each step scales idea distribution; humans consume the same ideas; human output becomes more same-y as well, with fewer novel ideas.

Frankly, I have serious doubts about the findings of this study. The LLM has been trained in every writing style known to humankind throughout history, from rigorous academic texts to street slang and philosophical musings. If we were to put this model in a writing competition with any human being in terms of 'diversity,' it would win by a landslide, not because I say so, but because the statistical range of the model far surpasses the capabilities of any individual mind, no matter how brilliant