Points 6 and 7 shows how well the students are intuiting LLMs. I think it ties very well into why they think progress from chatgpt 2 to chatgpt 5 was not a drastic improvement. After all, the first car Ford mass produced and current supercars are the same in principle, internal combustion engines for transportation. Engine design, technology development and increased complexity does not change that a car is a car. A Lamborghini is awesome in capability just as the latest LLMs are, but they are still what they are.
Wait, how does their intuition about AI desires (point 6) and the HuggingFace incident (point 7) have anything to do with why they think there's been only incremental progress from GPT-2 to GPT-5?
If anything it just feels like they don't understand how hard certain tasks are, so when AI can do those things, it doesn't mean anything to them. For example, in point 1, students' reaction to Claude Code isn't "wow, AI can code entire programs now", it's "wow, I had no idea software engineering was actually that easy."
I have experience with highschool students and I'd like to add one thing. There is a clear anti-AI attitude especially amongst brighter students because they see this as yet another clear attempt, in an education system already in chaos and staffed by underqualified teachers, to shift even more responsibility onto the pupils’ shoulders. Responsibility that ought not to be there in the first place.
The kids are alright! This list gives me great hope and it’s amusing to see the usual TESCREAL crowd lose it over at LessWrong over how they should be more worried about post-singularity runaway basilisks and less worried about enshittification, deskilling, and inequality.
Point 2.3 (unemployment feedback loop) is real and has occurred in history - it's called an economic depression. As in the Great Depression, but there are littler ones too.
It, like most of the economy, is a policy decision by Congress.
Points 6 and 7 shows how well the students are intuiting LLMs. I think it ties very well into why they think progress from chatgpt 2 to chatgpt 5 was not a drastic improvement. After all, the first car Ford mass produced and current supercars are the same in principle, internal combustion engines for transportation. Engine design, technology development and increased complexity does not change that a car is a car. A Lamborghini is awesome in capability just as the latest LLMs are, but they are still what they are.
Wait, how does their intuition about AI desires (point 6) and the HuggingFace incident (point 7) have anything to do with why they think there's been only incremental progress from GPT-2 to GPT-5?
If anything it just feels like they don't understand how hard certain tasks are, so when AI can do those things, it doesn't mean anything to them. For example, in point 1, students' reaction to Claude Code isn't "wow, AI can code entire programs now", it's "wow, I had no idea software engineering was actually that easy."
I have experience with highschool students and I'd like to add one thing. There is a clear anti-AI attitude especially amongst brighter students because they see this as yet another clear attempt, in an education system already in chaos and staffed by underqualified teachers, to shift even more responsibility onto the pupils’ shoulders. Responsibility that ought not to be there in the first place.
Attempt by who? Responsibility to do what? Not snark, just not following.
The kids are alright! This list gives me great hope and it’s amusing to see the usual TESCREAL crowd lose it over at LessWrong over how they should be more worried about post-singularity runaway basilisks and less worried about enshittification, deskilling, and inequality.
Point 2.3 (unemployment feedback loop) is real and has occurred in history - it's called an economic depression. As in the Great Depression, but there are littler ones too.
It, like most of the economy, is a policy decision by Congress.
Good article