5 ms·

Firstly, let us not conflate a full time professional YouTube personality with dedicated researchers, no matter how sexy the algorithm makes him. Secondly, we

Firstly, let us not conflate a full time professional YouTube personality with dedicated researchers, no matter how sexy the algorithm makes him. Secondly, we can demonstrate that this method is some AI hype nonsense because there exist hundreds of books on basic topics like mathematics all written for different learning styles. The LLM is not bringing anything to the table to make up for it's major shortcomings. Most importantly is that the human learning interface for children is far too vast for a simple model to attempt to interact with. Smells, hunger levels, nap phase, whatever. An LLM won't interface with it anymore than a pigeon can interface with a 24fps film. https://www.psychologytoday.com/gb/blog/what-shapes-film/201406/why-you-can-t-take-pigeon-the-movies https://www.psychologytoday.com/gb/blog/what-shapes-film/201...

moffkalast 3y ago
The video is 11 years old, it's not exactly something chasing hype. And while I'd be the first to argue against blindly following what some dude says on youtube, he does have two degrees, has worked as an actual teacher and has been a lifelong educator. Plus has a team to fact check anything he makes. Kinda gives him a moderate amount of credibility on this specific topic in my book. Yeah you probably can't teach first grade kids who don't know what a tablet is anything, even with a fully multimodal model. I'd imagine if there is any kind of transition towards this sort of approach it makes sense to start in university first, where the teachers are objectively the worst at teaching of all levels of education, and students the smartest. Then a level lower in high school once the systems work better, etc. Hell, one of the docents at my university lab would give anything for an automation system that would replace the lectures and practice classes he grudgingly has to do, lmao, I think the entry hurdle for adoption there is rather low.

moffkalast 3y ago

The video is 11 years old, it's not exactly something chasing hype. And while I'd be the first to argue against blindly following what some dude says on youtube, he does have two degrees, has worked as an actual teacher and has been a lifelong educator. Plus has a team to fact check anything he makes. Kinda gives him a moderate amount of credibility on this specific topic in my book. Yeah you probably can't teach first grade kids who don't know what a tablet is anything, even with a fully multimodal model. I'd imagine if there is any kind of transition towards this sort of approach it makes sense to start in university first, where the teachers are objectively the worst at teaching of all levels of education, and students the smartest. Then a level lower in high school once the systems work better, etc. Hell, one of the docents at my university lab would give anything for an automation system that would replace the lectures and practice classes he grudgingly has to do, lmao, I think the entry hurdle for adoption there is rather low.