Imo this is so overrated, the biggest difference is that children have 100% exposure and almost all language they come across is tailored exactly to the right level, just for them.

What would you do instead to make sure the student actually possesses the skills they are intended to have learned by the end of a program?

I'm not sure it's possible to force someone to learn who doesn't want to. From what I can tell, the article is basically saying that giving the students some form of agency and trust is a better way to motivate them to take it seriously than being a strict top-down disciplinarian. This fits with my experience (both from when I was a student and in my interactions with younger people as an adult), and I would expect that most people who have seriously evaluated this strategy would come to the same conclusion. It's not perfect, as some students may try to take advantage of things or will still phone it in, but the same thing happens with every other way of trying to engage with them.

My experience has been the complete opposite, a bit of pressure goes a long way. There are many people who need to know X or Y and just dont have the maturity or innate motivation to do it properly. This comes from the experience of a Dutch school system so perhaps its different in other countries.

A variety of performance assessments more similar to commercial pursuits.

Group projects with tangible artifacts, including finished prototypes that meet objectives. More emphasis on group projects. If AI accelerates productive development like with software, move the objectives up the ladder in complexity, or expectations.

Peer assessments and performance reviews like employment . This also helps prepare students for adult life.

If the subject matter is merchandisable, have the students operate an enterprise. My local high school has the students operate a food cart for example, and it opens to the public one weekend a month, otherwise open to students. Students are responsible for inventory, marketing, accounting, maintenance , customer service etc.

More verbal challenges . These can be operated by AI with human supervision while being recorded, with spot checks from supervisors.

Every diagnostic has a precision / recall curve and some fall through the cracks. But you have to shift your approach when old testing no longer becomes viable. Better that than to revert to the stone age of informatics.

> If AI accelerates productive development like with software, move the objectives up the ladder in complexity, or expectations.

The problem is that the purpose of group projects is (besides practicing programming) to facilitate learning together, splitting tasks, discussing approaches, presenting the outcome. Doing these requires understanding what's going on, and if you just vibe code everything you don't have enough knowledge to experience the soft-skill parts of the work.

And especially China. And Japan.

If you had an llm that could accurately predict when a claim is uncertain it would be very popular, I think. I would pay for that kind of reliability tbh

This would break reality. There’s some underlying physical law that prevents the existence of any algorithm of truth.

> There’s some underlying physical law that prevents the existence of any algorithm of truth

Haven't heard about that law, but seems unlikely we can come up with ("discover") any sort of law that uses a concept ("truth") humans can't even agree what it means, and that's not for a lack of trying, we've been trying to figure it out for millenniums already with no end in sight.

If you accept certain axioms a priori, it’s fine. If you simply let the machine intelligence take it for granted that induction works because nature is uniform and give it some way to test its predictions, it would have all the building blocks it needs to reason out a lot of very useful information. Which as the parent comment points out, people would absolutely pay a lot of money for.

If I am reading you correctly, that ends in testing, or essentially doing science.

That wouldn't satisfy the requirement to look at a claim and be able to tell if a claim was uncertainm roughly speaking.

You dont have to literally send a null token. Train it to generate text that summarizes the evidence that is there but the uncertainty of the final answer to a prompt.

Has nothing to do with size, see Japan and China.

What does it have to do with then?

My understanding japan has a surprisingly private train system, and China fits the model of "works when it's constantly expanding"

Japan is mostly using exactly the model I suggested.

I think they mean that a phd doesn't mean much relatively speaking, since everyone around has one so it's less impressive and you're less of an expert when everyone around you is knowledgeable in the same domain.

Nonsense, some of my friends are lawyers and they're able to give you consistent interpretations on why they think about a certain aspect of a law a certain way. The whole thing is that they work with this the entire time, so they have a really consistent 'head model' of how things work and why and how considerations should be weighted/ordered/whatever. LLMs just do not have this, there's no consistent underlying reasoning (the 'reasoning' traces in LLMs are really inconsistent)

Who are you in this comparison?

Completely agree, I get that this is a stepping stone for future, more reliable robots but I found the demonstration underwhelming.