I was hoping AI would take us there. All the kids have computers. A guided LLM walks them through videos and concepts. tests them, lets them advance at their own pace.
I wanted to zoom through the lessons, learn everything, but I didn’t want the awful college-prep bullshit or the high scrutiny of my essays. Did I understand the work? Yes, and that’s enough. I showed up to every class, knew what they asked, and aced tests.
How cool would it be for every kid to be able to ask unlimited questions and be given different ways to understand a problem and its corresponding answer until they understood them properly?
Set the text books and workbooks, guardrail the conversations. Double-check for incorrect assertions.
Let them go and seek out on their own whatever subject they wanted at whatever time they wanted.
Personally, if AI wasn’t so destructive, or blatantly stolen as it was, and they changed basically everything about it to make it more reliable and stuff, I agree. I do think eventually something that is called AI will do that, but it won’t be the LLM tech we have today that’s for sure, it will be some other form of tech.
I would love if they had something like you see in the futuristic movies where theres a hologram teacher, could have every kid have their own. I just don’t forsee it happening reliably with our current tech.
Reliable teaching would be straightforward, but not cheap to set up. You wouldn’t want the models used to have the entire sum of the internet trained. Just verified clean data.
You could start with textbooks, peer-reviewed studies, IC data from teachers, then add 20% reviewed, distilled content from a fully trained LLM. Set the whole thing up on rails. Curriculum, pacing, and testing would all be classical programming, architected by people, using a more widely trained LLM to sort out language and edge cases. New edge cases get human review. Set up a quorum of models that get hit on every request to sus-out errors. Backend is the source of truth, front end is not allowed to countermand the backend, and that’s tracked by another quorum.
I think the biggest problem isn’t doing it, the biggest problem is not destroying society to get it done and not allowing society to take it over with an agenda. We can’t have the teachers disappear. It’s ripe for corpo/government manipulation. For something like this to work, you can’t have capitalism driving profits. You can’t have cults indoctrinating. It would all need to be very post Bell Riots Star Trek, and we KNOW that’ll never happen.
I was hoping AI would take us there. All the kids have computers. A guided LLM walks them through videos and concepts. tests them, lets them advance at their own pace.
I wanted to zoom through the lessons, learn everything, but I didn’t want the awful college-prep bullshit or the high scrutiny of my essays. Did I understand the work? Yes, and that’s enough. I showed up to every class, knew what they asked, and aced tests.
How cool would it be for every kid to be able to ask unlimited questions and be given different ways to understand a problem and its corresponding answer until they understood them properly?
Set the text books and workbooks, guardrail the conversations. Double-check for incorrect assertions.
Let them go and seek out on their own whatever subject they wanted at whatever time they wanted.
Personally, if AI wasn’t so destructive, or blatantly stolen as it was, and they changed basically everything about it to make it more reliable and stuff, I agree. I do think eventually something that is called AI will do that, but it won’t be the LLM tech we have today that’s for sure, it will be some other form of tech.
I would love if they had something like you see in the futuristic movies where theres a hologram teacher, could have every kid have their own. I just don’t forsee it happening reliably with our current tech.
Reliable teaching would be straightforward, but not cheap to set up. You wouldn’t want the models used to have the entire sum of the internet trained. Just verified clean data.
You could start with textbooks, peer-reviewed studies, IC data from teachers, then add 20% reviewed, distilled content from a fully trained LLM. Set the whole thing up on rails. Curriculum, pacing, and testing would all be classical programming, architected by people, using a more widely trained LLM to sort out language and edge cases. New edge cases get human review. Set up a quorum of models that get hit on every request to sus-out errors. Backend is the source of truth, front end is not allowed to countermand the backend, and that’s tracked by another quorum.
I think the biggest problem isn’t doing it, the biggest problem is not destroying society to get it done and not allowing society to take it over with an agenda. We can’t have the teachers disappear. It’s ripe for corpo/government manipulation. For something like this to work, you can’t have capitalism driving profits. You can’t have cults indoctrinating. It would all need to be very post Bell Riots Star Trek, and we KNOW that’ll never happen.