Khan Academy integrates GPT-4 as every student’s customized tutor • Hacker News Board

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#191

Post by falcor84 » Wed, Mar 15, 2023, 2:25 PM UTC

pnt12 wrote: post #166

superfrank wrote: post #126

Earlier quoted context omitted.

Is there a name for this fallacy (is it a fallacy?). I feel like we see the same issue with self driving cars crash rates. People are hesitant to adopt because they aren't perfect, but ignore that it's better than the status quo.

I'll don't think it's a fallacy. People are more predictable and can be accountable for their actions, and even admit ignorance, eg answering "I don't know ask someone else". It's worth discussing if we want AIs replacing them - maybe the scalability is better, but these are usually worse.

> People are more predictable and can be accountable for their actions...

I would disagree. I think that at this stage, ChatGPT is very predictable, and we have excellent statistics on how likely it is to respond correctly to a particular type of question. I would have much less predictability about any particular human teacher, who might have an especially good or bad day.

Also, ChatGPT can be much more accountable, in the sense that when issues pop up, a filter can be implemented to prevent these particular issues for arising again. I don't see how humans are better in this regard - even if they are "accountable" in some vague sense, it can take many months to get rid of an underperforming teacher, and can take decades to change something across the system.

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#192

Post by falcor84 » Wed, Mar 15, 2023, 2:32 PM UTC

yamtaddle wrote: post #156

worrycue wrote: post #142

Earlier quoted context omitted.

There is being misinformed and making a mistake when reasoning. Then there is making s** up. Most humans have the decency to not do the latter while AI (at the moment) freely makes up complete lies when it doesn't have the answer instead of just admitting it doesn't know.

At least misleading instruction from teachers tends to be of a sort that others also experience, so can later be addressed en masse (e.g. mythologized history, poor-but-common explanations for things like how airplane wings work). Or spreading the same political propaganda that can be found all around, so is pretty easy to spot. It's not usually that you ask them a question they can't answer, and they just confidentl…

I suppose that this is true. But at least with an AI under your control (as a Khan Academy admin), you can run a query to analyze any unexpected decrease in student performance, find correlations with changes in the keywords that these students received from the AI tutor, isolate the particular misleading AI outputs by running some (possibly automated) experiments, and then once identified, fix the problem systemically by prompt engineering, fine-tuning or output filtering.

What's the analog of that with human teachers, and how long would it take to implement a solution there? Even if you had all the budget in the world, how would you go about making all teachers stop misleading students about how airplane lift works?

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#193

Post by falcor84 » Wed, Mar 15, 2023, 2:33 PM UTC

itronitron wrote: post #118

worrycue wrote: post #67

Am I the only one somewhat alarmed that we are choosing to rely on AIs that are vulnerable to hallucinations so much? Personally I rather have a more limited but reliable tool than a more powerful but unreliable one.

There are two indicators that this is PR bullshit produced by whoever is trying to capitalize their investment, 1) is the use of the marketing term 'AI' or 'AI-powered' and 2) use of the 'think of the children' trope.

'AI' is a marketing term? I don't see how that's more than saying that 'cloud' is a marketing term - while being useful in marketing, it's just an industry term.

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#194

Post by falcor84 » Wed, Mar 15, 2023, 2:42 PM UTC

sophiabits wrote: post #79

alach11 wrote: post #57

Earlier quoted context omitted.

My mind went to the same place. Knowing that "the average tutored student was above 98% of the students in the control class", imagine how transformative it will be to lift the average student to the 98th percentile! This may be the most transformative change to education since the invention of public schooling.

Every now and then I think about Marc Andreessen’s “software is eating the world” article. He was extremely excited about investing in EdTech and HealthTech software companies at the time—for obvious reasons! But EdTech hasn’t really delivered much value. The modern LMS feels to me like a reinvention of Facebook groups in many ways… LLMs feel like something that _could_ very well end up eating education, and that’s a…

> The modern LMS feels to me like a reinvention of Facebook groups in many ways

I agree with your sentiment that there's a lot more potential value to be derived from EdTech than we've seen so far, but even if LMS's are "just" like Facebook groups, that's already the equivalent of hundreds of billions of dollars in economic value.

When I was a kid, if I was sick for a few days, I had to manually copy my friend's notes after I returned to school, and would still feel confused. Now, with an LMS, I have access to all the learning content from home, for my own class, and for resources from similar classes taught by amazing teachers from across the world, at any modality I prefer. This on its own is incredible value. (although with the caveat that being sick is no longer the excuse to rest and relax as it used to be)

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#195

Post by falcor84 » Wed, Mar 15, 2023, 2:46 PM UTC

jamestimmins wrote: post #14

Highly recommend the Neal Stephenson book The Diamond Age: Or, A Young Lady's Illustrated Primer , for an interesting exploration of a custom AI tutor for each student. In that book, students have a "magic book" that teaches them lessons in a story form while encouraging certain life paths. It's pretty fascinating to consider the implications and whether that's something we'll want, bc it may soon be possible.

Having heard many recommendations, I tried reading that book more than once, but have real issues with Stephenson's writing style there. Could you please recommend an overview of the qualities of the AI tutor that I could go over instead?

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#196

Post by falcor84 » Wed, Mar 15, 2023, 2:49 PM UTC

gsatic wrote: post #44

So untested and using kids as guinea pigs again. Well done. Keep it up. Just a never ending mindless need to make big claims constantly.

They literally announced that they're testing this on a small sample of kids whose parents opt-in. I don't see any better approach, seeing how the sentiments of "untested" and "using kids as guinea pigs" are contradictory.

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#197

Post by aitchnyu » Wed, Mar 15, 2023, 2:52 PM UTC

sebasvisser wrote: post #174

davesque wrote: post #141

Earlier quoted context omitted.

Yes, but human teachers can eventually come to understand why they made a mistake. And they probably are more able to communicate how confident they are in their answers. I think that's because human teachers have a mental model that they're working from. On the other hand, we have no idea if transformer models are working from anything like a "mental model." Just because humans make mistakes (or even as many or more…

Human teachers realising their mistakes and learning from it… how long ago was it that you were in an average school?

One teacher claimed he watched Animal Planet showing ostriches burying their head in the sand and humans casually slaughtering them.

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#198

Post by falcor84 » Wed, Mar 15, 2023, 2:52 PM UTC

8note wrote: post #152

How do you monitor for ai correctness on these topics? For gpt-3, that's been the user, but a student roughly has to take the ai at face value. Some secondary ai double checking the output? What's the loop like for saying oops to the student?

In this case, as I understand it, the idea is that the existing Khan Academy system generates a random question, the AI helps the student solve it, and then the student provides the answer to the original system, which then validates it.

I would also assume that Khan Academy could inject the true answer into the GPT prompt and instruct the AI to hide it from the user, but rely on it in its explanation, in order to reduce the risk of misleading outputs.

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#199

Post by kobalsky » Wed, Mar 15, 2023, 3:44 PM UTC

worrycue wrote: post #67

Am I the only one somewhat alarmed that we are choosing to rely on AIs that are vulnerable to hallucinations so much? Personally I rather have a more limited but reliable tool than a more powerful but unreliable one.

If you don't know how they are using GPT-4 it's not fair to say it will hallucinate.

As far as I understand the preferred way to use LLMs nowadays for domain specific information retrieval is through embeddings that insert the related context in the prompt. GPT-4 is specially good for this since they increased the prompt size almost by an order of magnitude.

This means that the model can be given a very specific task: to extract information from the context or avoid providing an answer at all.

The answer doesn't rely on the neural memory of the model, since it doesn't need to store information, just understand the task, and they are really good at that.

Re: Khan Academy integrates GPT-4 as every student’s customized tutor

#200

Post by alach11 » Wed, Mar 15, 2023, 4:43 PM UTC

recuter wrote: post #159

mullingitover wrote: post #39

I wonder if this kind of intelligent tutoring could be the answer to Bloom's Two Sigma Problem. The limiting factor with that problem was that not everyone can afford a personal tutor. Having an AI tutor that can breeze through the SAT seems like it should give every student a major boost.

A tutor doesn't just imply parents who can afford one but also who care enough to do so in the first place and see the value of an education. I think you'd find that even by the sat-prep age the actual limiting factor for many (most?) of the tutor-less is a fate-accompli attitude towards their path in life. Might as well turn on the chat bots at the cradle. Ah, and make sure everybody gets the same quality AI to pare…

Right... but the whole point Bloom's experiment is that if you give every kid a tutor, they perform (on average) at the 98th percentile. Not just the rich kids who can afford tutors.

I do think there are valid concerns that this technology (like old-school human tutoring) could be disproportionately available to already-privileged children. This could potentially widen the achievement gap. But it seems way more likely that we can get poor kids a laptop to tutor them vs. getting them a dedicated human tutor.