Live data from Hacker News

AI Homework

stratechery.com

41–50 of 296 posts

Re: AI Homework

#41
What people are missing here is that you can setup feedback loops of models to get the "correct" content or whatever you want.

For example, If you get too much NSFW content put a NSFW detection step on top of GTP ("Is this response offensive") and have that model respond to GTP with "please repeat but with fewer bad words".

For accuracy you can add a Wikipedia feedback loop. "Does this response match information on Wikipedia?" if NO ask it to try again.

Think of these models as transistors or basic logic gates. The real power comes out when you link them together into more advanced logic like a Instruction Pipeline circut.

Re: AI Homework

#42
I’m surprised by the statement that cost is a limiting factor. If the model produces correct high quality output, 750 words for 2 cents is a bargain! Do you know how long it would take me to write 750 words??? Even if I only spend 10 minutes it’s a no-brained to use this tool to help.

Re: AI Homework

#43

Earlier quoted context omitted.

Over the weekend Twitter was 30% ChatGPT screenshots and 20% engineers frothing about how ChatGPT would replace programming, article writing, even Google Search. Such engineers must be juniors, or have persisted in their life with a totally wrong idea of how engineering is done and engineering is built. I can't imagine thinking that one could build software systems or essay arguments by trusting an AI system you don'…

Couldn't this help a 10x engineer become a 100x engineer? The secret is still going to be in understanding what to ask for and ensuring that what gets spit out works. It's just going to be so much faster. Is there anyplace in the world for a .5x developer or even a 1x developer in a chatgpt world?

But this sort of change is happening all the time in all sorts of different domains. There probably isn't much of a role for a web developer who just knows some basic HTML like there was in the late 90s. And you can apply this to a lot of basic wring, transcriptions, etc.--at least at western pay scales.

Re: AI Homework

#44
post #8

I love talking to GPT, but I've tried using it for some serious stuff (research, biz-dev) and less than serious stuff (Dungeons & Dragons) and the magic is still there but it has obvious blind spots which might be solved in the future but we're talking about the present. I am wholly amused how everyone got on that dude from Google that accused the AI of being sentient and Google about not caring, and now you have hug…

There's a huge gap between "The breadth of training this AI has and the depth of its processing is impressive" and "This script is sentient". The latter is a _far_ broader (and unsubstantiated) claim.

Re: AI Homework

#45

ChatGPT looks awesome, I'm really interested to see how it does with more niche topics. E.g. give it high school level, undergrad level, grad level, and research level math topics to generate notes for and see how it does for each one

Badly. It can't do math, to an extent which is honestly surprising given how well it produces code. I suspect that this will improve significantly in the next couple of years, but I kind of hope I am wrong. I am a mathematician by training, and I have my own romanticism.

For example, I tried a bunch of times to get it to give a proof of the infinitude of the primes. This should be repeated many times in its training data, and it clearly is given it's responses, but it always goes off the rails around the third sentence. Here's an example:

> One way to prove that there are infinitely many prime numbers is to use the contradiction method. Assume for the sake of contradiction that there are only finitely many prime numbers, say, p₁, p₂, ..., pₙ. Then we can form a new number N by multiplying all of these prime numbers together and adding 1: N = p₁ * p₂ * ... * pₙ + 1. Because N is greater than any of the prime numbers p₁, p₂, ..., pₙ, it cannot be divisible by any of them. Furthermore, N is not 1, so it cannot be a composite number (i.e., a number that can be written as the product of two smaller numbers). Therefore, N must be a prime number, contradicting the assumption that there are only finitely many prime numbers. This contradiction shows that our assumption must be false, and therefore, there must be infinitely many prime numbers.

Re: AI Homework

#46
post #16

"The solution will be to start with Internet assumptions, which means abundance, and choosing Locke and Montesquieu over Hobbes: instead of insisting on top-down control of information, embrace abundance, and entrust individuals to figure it out." While this sounds good, it depends on a crucial assumption: that there are places on the Internet where people can go to find content that is not generated by an AI, but by…

> Just one problem: what happens when all of the content on the Internet (or at least the vast majority of it) is generated by AI? The vast majority of the content of the internet is junk now, how do people deal with that? They have specific known sources that they trust to faithfully relay primary works and/or to provide trustworthy analysis.

I was going to say books, then my colleague wrote an entire book using ChatGPT at work today.

Re: AI Homework

#47
post #15
post #7

There has been a lot of "well, we had a good run" comments on ChatGPT threads from engineers lately. I get where this sentiment is coming from, but I also think that the article paints a good picture of how we must "pivot" as a species to avoid faceplanting completely. Zero Trust Homework sounds like a strategy we will be forced to implement: the system will frequently give the wrong answers (and not just on accident…

I can imagine somebody knowing the answer being able to properly verify the output. But a student which is supposed to learn something form that homework? We're missing the entire point of homework here.

Actually this is interesting framing given the misinformation pandemic on the internet today. Suppose students are taught material and then given a test in the form of a mix of correct and subtly incorrect assertions that they need to filter through. Might help them learn to distinguish information from misinformation as well?

Re: AI Homework

#48
post #23
post #15

Earlier quoted context omitted.

I can imagine somebody knowing the answer being able to properly verify the output. But a student which is supposed to learn something form that homework? We're missing the entire point of homework here.

Well, they would need to know something about the topic. The easiest way for them to do this would be to read about it. When picking what to read they would need to be able to estimate if it contained valid information. The meta-skills that would be developed would be: * Understanding the quality of sources * Understanding the different between primary and secondary sources. * Establishing chains of evidence and trac…

Someone was also demonstrating over the weekend that literally cut and pasted ChatGPT text output is easily detected as LLM output with near certainty. https://huggingface.co/openai-detector/

That said, adult me could probably use ChatGPT to stitch together a 1000 word high school paper pretty easily (and probably evade plagiarism/LLM detectors). But, as you say, I could probably get a lot of the way there by other means as well and putting words on the page is mostly not the hard part anyway.

Re: AI Homework

#49
Zero Trust Homework is a fascinating idea, as described in the article, but I don't know how many teachers will adapt because they'll just end up confusing themselves. When you read too many lies, you might start forgetting the truth yourself.

Instead, what I think there will be a much greater short term market for, is a system that will run the teacher's homework assignment (with variations) against GPT-3, ChatGPT, and similar, and will cache the results. Then, when students submit the homework in their LMS, it will automatically compare the cached results for repeated sentences and paragraphs and flag the submission if it matches the AI results too closely. I expect TurnItIn will do something like this eventually, but what if somebody beats them to the punch?

If I had a million dollars, I'd have it built overnight and then marketed to every college, ASAP. I'd tell every sales agent to have teachers and college executives try TextGPT, panic, and sign immediately. I'd sign a million contracts before TurnItIn knew what hit them. Then, once suddenly and so quickly entrenched, I'd extend it into a TurnItIn competitor.

Re: AI Homework

#50
post #13
post #9

Earlier quoted context omitted.

If this were a test question, the response given would be marked wrong, likely with no partial credit awarded. It's that egregiously wrong, even if the attribution is perhaps understandable.

Well thankfully I haven't been in school for over a decade at this point, so rarely (never?) encounter these hard-cutoff no-second-chances trivia situations. I operate in the real world, where continually making something that's not-quite-correct then refining it and fixing the errors is how basically everything is accomplished.

> not-quite-correct then refining it and fixing the errors is how basically everything is accomplished.

Not-quite-correct is a wordy misrepresentation when something is plainly incorrect.

In many fields being confidently wrong is how you get fucked over hard.

That ChatGPT will gleefully fabricate references like a schizophrenic is just more icing on this shit cake.

Post reply on HN