I get these kinds of things all the time at work. I’m a writer, and someone once sent me a document to brief me on an article I had to write. One of the topics in the briefing mentioned a concept I’d never heard of (and the article was about a subject I actually know). I Googled the term, checked official sources … nothing, it just didn’t make sense. So I asked the person who wrote the briefing what it meant, and the response was: “I don’t know, I asked ChatGPT to write it for me LOL”.

Same, I use it to put me down research paths. I don’t take anything it tells me at face value, but often it will introduce me to ideas in a particular field which I can then independently research by looking up on kagi.

Instead of saying “write me some code which will generate a series of caverns in a videogame”, I ask “what are 5 common procedural level generation algorithms, and give me a brief synopsis of them”, then I can take each one of those and look them up

$100 billion and the electricity consumption of France seems a tad pricey to save a few minutes looking in a book…

I recently read that LLMs are effective for improving learning outcomes. When I read one of the meta studies, however, it seemed that many of the benefits were indirect: LLMs improved accessibility by allowing teachers to quickly tailor lessons to individual students, for example. It also seems that some students ask questions more freely and without embarrassment when chatting with an LLM, which can improve learning for those students - and this aligns with what you mention in your post. I personally have withheld follow-up questions in lectures because I didn’t want to look foolish or reveal my imperfect understanding of the topic, so I can see how an LLM could help me that way.

What the studies did not (yet) examine was whether the speed and ease of learning with LLMs were somehow detrimental to, say, retention. Sure, I can save time studying for an exam/technical interview with an LLM, but will I remember what I learned in 6 months? For some learning tasks, the long struggle is essential to a good understanding and retention (for example, writing your own code implementation of an algorithm vs. reading someone else’s). Will my reliance on AI somehow damage my ability to learn in some circumstances? I think that LLMs might be like powered exoskeletons for the mind - the operator slowly wastes away from lack of exercise.

It seems like a paradox, but learning “more, faster” might be worse in the long run.

If it’s a topic that has been heavily discussed on the internet or in literature, LLMs can have good conversations about it. Take it all with a grain of salt because it will regurgitate common bad arguments as well as good ones, but if you challenge it, you can get it to argue against its own previous statements.

It doesn’t handle things that are in flux very well. Or things that require very specific consistency. It’s a probabilistic model where it looks at existing tokens and predicts what the next one is most likely to be, so questions about specific versions of something might result in a response specific to that version or it might end up weighing other tokens more than the version or maybe even start treating it all like pseudocode, where descriptive language plays a bigger role than what specifically exists.

I agree. I was almost skipping it because of the title, but the article is nuanced and has some very good reflections on topics other that AI. Every technical progress is a tradeoff. The article mentions cars to get to the grocery store and how there are advantages in walking that we give up when always using a car. Are cars in general a stupid and useless technology? No, but we need to be aware of where the tradeoffs are. And eventually most of these tradeoffs are economic in nature.

By industrializing the production of carpets we might have lost some of our collective ability to produce those hand-made masterpieces of old, but we get to buy ok-looking carpets for cheap.

By reducing and industrializing the production of text content, our mastery of language is declining, but we get to read a lot of not-very-good content for free. This pre-dates AI btw, as can be seen by standardized tests in schools everywhere.

The new thing about GenAI, though is that it upends the promise that technology was going to do the grueling, boring work for us and free up time for us to do the creative things that give us joy. I feel the roles have reversed: even when I have to write an email or a piece of coding, AI does the creative piece and I’m the glorified proofreader and corrector.

I think the author was quite honest about the weak points in his thesis, by drawing comparisons with cars, and even with writing. Cars come at great cost to the environment, to social contact, and to the health of those who rely on them. And maybe writing came at great cost to our mental capabilities though we’ve largely stopped counting the cost by now. But both of these things have enabled human beings to do more, individually and collectively. What we lost was outweighed by what we gained. If AI enables us to achieve more, is it fair to say it’s making us stupid? Or are we just shifting our mental capabilities, neglecting some faculties while building others, to make best use of the new tool? It’s early days for AI, but historically, cognitive offloading has enhanced human potential enormously.

Proof that it’s already too late ☝️

I was annoyed that it wasn’t over her mouth to implant the egg.

(picking up phone) Hello this is Sherlock speaking

You’re speedrunning Danning-Kruger with an impressive force

I use LLM’s to help with math/science/coding, and the thing it screws up the most seems to be simple math (typically units/conversion issues) so I would be weary about gleaning financial advice from a chatbot.

You don’t need AI to do that…