Comment by noduerme | Hacker News Reader

Comment by noduerme | Hacker News Reader

Parent Full thread

Can you name anything you've learned from an LLM? Start your essay with what you've learned from AI as your opening paragraph. Write by hand as if you had been educated to do so. List one or more specific instances in which an LLM gave you knowledge which you would take forward and use productively, as your own integrated knowledge, in such a way that you would be consistently able to apply it to novel problems and circumstances on your own.

I won't down vote you, because you simply misunderstood me. Of course I'm using this incredible technology to accelerate writing code.

If you never learned the value of learning itself, for its own sake, you have everything to gain from a world in which skills like learning and thinking for yourself are useless. I can hardly blame you for that.

But as someone who has learned by teaching others, people who in turn will learn by teaching others, I don't find your belief in someone else doing the work for you to be particularly sustainable. Or convincing. Regardless of whether I think it's contemptible, lazy, or damaging to your mind, which isn't my problem.

> Can you name anything you've learned from an LLM?

Not GP, but here’s some of the things I have learned in the last couple weeks from LLMs while asking LLMs to review parts of my code and documentation. I wrote the code and documentation without LLMs.

Linux user permissions is a fake because by design, the owner has permissions to write these 3 bits. A good way to reliably deploy a readonly file is owner=root, assign the read permission to the group.

chown command follows links by default, -h switch disables that.

In Linux, resetting environment variable in a long-lived server process doesn’t erase much data because /proc/{pid}/environ is a snapshot taken on process startup.

If my server crashes (or manually killed by a root to test things on staging VM) and Alpine’s supervise-daemon restarts it, the start_pre() function from the OpenRC service script won’t be called again. A fix is use OpenRC to run a launcher shell script as root, in that script manually downgrade the account with `exec su-exec svc:svc /the/server/elf`

This last one is irrelevant to my use cases but still interesting. AES GCM may fail catastrophically when different message encrypted with same password and nonce. For 12 bytes of random nonce, the stuff is only safe up to approximately 4E+9 messages using the same password, beyond that birthday paradox ruins things.

Absolutely, this week I learnt about:

- Kalman Filters

- Bayesian Tracking

- Hybrid Logical Clocks

I was able to ask questions and understand the current state of the art.

How do you know the explanations were correct?

How do you know any explanation is correct from any source? That is not unique to LLMs.

You find sources written by experts and vetted by peers.

This is not very hard, and we all used to know how to do this not 3 years ago.

The LLM is citing it's sources and I follow up on them. Your attempt at a gotcha is interesting in that I think it says more about your relationship with AI than mine.

This is the story I hear over and over again when an expert evaluates using AI for an introductory course inside their own field. The thing is full of mistakes, bad or wrong information, and unhelpful advises.

https://www.youtube.com/watch?v=rdj_Fm4XM60

If a person writes a textbook, a tutorial, a blog post etc. which is that bad it won‘t be recommended by peers and a beginner will either not know it exists, or find see the bad reviews from peers. The beginner will then look for a different textbook, tutorial, blog post etc.

When self learning with AI, you don‘t get that. The best advice a peer can give you, is simply, do not use AI to learn stuff you don‘t already know.

I'll wank my way and you wank yours.

Reply on news.ycombinator.com