"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
Students will prioritize learning when learning is the goal.
The goal of modern school systems is day care, and then it's to force children to get used to a 40hr work week. The third and least important excuse for school is to get a degree, which is now the bare minimum to be allowed to enter the workforce.
None of these things have to do with learning the source materials.
Before AI, rich kids hired poor smart kids to write all their papers for them.
Real classes had exams where you could bring all your notes and textbooks and you still failed.
And when they got out into the field with their dubious degrees, they had daddy’s connections to land a gig where they never needed to actually use the knowledge they pretended to have.
Not sure what we’re gonna do now that problem-solving positions are about to be filled by people whose only training is to produce incremental remixes of old solutions to well-documented problems.
Particularly with climate change promising to invalidate some fundamental assumptions baked into the literature of a significant chunk of our most important professions.
If understanding first principles means you move slower, look less polished, and are at odds with the majority of the paid staff in the field, then you’re not gonna make it.
And there’s a feedback loop, too. If it’s easiest to produce stuff that agrees with everything that came before, then that’s all we’ll make, and that’s all we’ll train new AIs on, so that’s what will continue to be easiest to produce.
College builds civilization by staging collaborations of people dedicated to integrating threads of past information with one another to generate and share new information. Relying on LLMs to perform said integrations lowers the quality of the generated information unless you also have the same LLMs presenting their research at conferences, performing peer review, and other collaborative aspects of academia.
As bad as it sounds, seeing someone do some "vibe doctoring" sounds hilarious.
Of course, this only applies if you're not the patient.
One of these days, we're going to find out some student somewhere deepfaked themself, used an LLM or whatever, and connected it all to Zoom or whatever remote learning platform to try to fully automate obtaining a degree.
There have been stories about people outsourcing their own jobs to China and collecting (part of, I guess) the paycheck. I can't imagine someone isn't already working on doing something similar with deepfakes and/or LLMs in either a work or academic context.
LLMs are genuinely a fantastic learning tool. You just have to use it as a learning tool, not a learning replacement...
Nah, they hallucinate waaaayyy too much
They hallucinate too much to be used as a learning replacement. The hallucinations are completely fine when you're using it as a learning tool because when used correctly, you verify everything externally. I've used it to breach some difficult stuff I was really struggling with, like the Rust programming language. Used it to help me get started on it the right way, then used it to identify gaps in my knowledge. Been writing Rust at a pretty high level for 3 years since with no assistance from AI.
