Zum Inhalt der Seite gehen


I don't think that "give beginners LLMs to get them started so they can actually learn to do it later" checks out.

(Original title: But will they?)

tante.cc/2025/07/21/but-will-t…

tante hat dies geteilt.

Als Antwort auf tante

great essay

very strong argument that is completely agnostic to whether LLMs are actually capable of what they advertise, very difficult to hand-wave this sort of rhetoric

Als Antwort auf tante

"That’s on us for shaping our education system into something not about intellectual curiosity and development of skills but into a form of certification generator that might get you a well-paid job."

💯

@rebeccawatson made a similar point here.

skepchick.org/2025/07/study-ch…

Als Antwort auf tante

i would even argue that the exact opposite is more reasonable because then you actually have the skills to judge the output. it’s the whole you have to know the rules to break them thing.
Als Antwort auf tante

there is a way to use them to learn effectively. But people have to know how to use them this way. If they don't, they won't. This is the root problem. People don't learn how to interact properly and effectively with (this) tech.

Senior people in the field probably figure it out. Junior people might. Laypeople are lost.

Als Antwort auf tante

they may have changed strategy since then, but as of when I left a certain college the CS department’s philosophy was “you have to do the introductory courses/learn the theory by hand. Once you’ve proven you know the fundamentals, you’re allowed to use GenAI, but you have to disclose it, and you’re still responsible for making sure your code passes all the test cases”.

Which is, as others have pointed out, the exact opposite of what the panelist proposes.

And is something I’ve ranted on before:

public.milcyber.org/activities…

Als Antwort auf tante

yes, every cheater in every test sat down and learned the stuff 4 months later in earnest when I was school and university. 🙄🙄🙄
Als Antwort auf tante

"There is no value in something that is only sweetness and gentleness. There is no meaning to a live that's just given to you. Even if it's difficult and tough, the things you build up through hard work in reality is what really makes you happy. Please remember that."

Quote from Arifureta, season 3, ep 6, 21:15. This is a trashy, nasty anime—one of those with tits and everything. Why does it know something the aipologists don't?

Als Antwort auf tante

I am preparing for my master's degree which starts in autumn, and for that I looked into the science of learning. I found one paper about study techniques, and testing yourself on studied material is hard but the most effective method to understand and move knowledge to the long-term memory.
Using LLMs to solve problems will rob you of this learning effect. The paper on learning: cognitiveresearchjournal.sprin…
Als Antwort auf tante

@tante@nettime.org Well put. In a way, this reminds me of my personal experience with YouTube tutorials and essays. Consuming them let's you use the lingo and fake understanding, but actually trying the things that one supposedly now understands leads to frustration and disillusionment, at least in my experience.

I've really come to appreciate the word practice a lot more over the last two decades.

Dieser Beitrag wurde bearbeitet. (5 Tage her)
Als Antwort auf tante

The problem with giving a todler a 50 cal machine gun is it takes a lot of strength to pull the trigger...

As soon as someone says, you can't do that.. someone is going to try and prove them wrong, by that time it's too late.

When do you think, we will see the great AI disaster of '26?

Als Antwort auf tante

I agree with your analysis that llm's impede learning, and of course universities should try to make people do real practice, but I don't think @mitsuhiko says anything against that (in the essay at least), but rather "people will vibe code anyway, we should be open and invite them so they eventually really learn"?