this is a fantastic idea. Got me thinking about ways to engage the people at my company with ChatGPT. I could imagine doing something similar as a training exercise that gets people familiar with the tech.
I like this idea. I think it would be good for students and teachers alike. Embracing AI in this way provides a good opportunity for students to learn how to best use the tech, lessons like the importance of asking the right questions and checking sources. And if teachers don't utilize these tools, they won't be able to take advantage of what AI has to offer or prevent outcomes they don't like, such as cheating on exams.
'Autodidact' is doing some heavy lifting here. Autodidact means self-teaching, but actually, everybody is self-teaching when they actually care about learning something. Practically everyone has gone on youtube to watch an instructional video on how to do something.
Autodidacts are more self-motivatatng than self teaching ("auto-proponits" maybe), and so will tend to figure out how to make best use of whatever tools become available to accomplish their learning goals.
But for most students (at any level of cognitive ability) it's motivation, not tools or techniques, that is the key issue. The problem is not lack of tools, access, or resources, which have been around for a long time in powerful forms and at low prices which would have astounded preceding generations. The problem is shaping situations and incentives to nudge the marginally un-motivated student to *want* to learn.
My hunch is that at this point we run into issues of hard-wired social psychology that make it difficult to substitute those social influences with computer tools. For example, it is instinctively easier for a member of 'the audience' to pay attention to some focal 'performance' when that member subconsciously picks up the subtle cues that the people around him are also paying close attention. It is easier to do what one imagines a teacher or professor or mentor wants (or what would impress them as an opportunity to show off) when one subconsciously calculates that affilitating with an authority figure or high-status individual would likely have personal benefits down the line.
Chatbots don't have status, can't do anything for you, don't make you cool or sexy when you tell people you use them. Interacting with a chatbot - in what is bound to be a typically solitary fashion - doesn't provide those social cues or incentives, and thus, doesn't help coach or motivate (at least, not yet, maybe soon!)
So, the autodidacts / autoproponits will indeed derive a lot of advantage from this new powerful tool, just like they did with all the previous powerful tools. But the other students will just use it like a search engine as an even faster way to get the answers to the homework or test or lines for the essay.
Perversely, I suspect that being even more able to outsource research, learning, and now even impressive composition to the chatbots will actually undermine the retention of actual skills or knowledge, just like people who were once able to use paper maps to navigate around a city found that they couldn't do it anymore after years of relying on smartphone GPS apps.
Surprisingly, I'm pretty sure most of this is not quite right. Yes, the self-teachers will use the tools the best, but I'm pretty sure many medium poor students will get turned on by the ability to get understandable answers to their questions, helping them more quickly understand the world.
Yes, many or most students will use it to get homework done as quickly (and with as little retained learning?) as they can. And even more will remember less - of the uninteresting stuff, usually what they have that they don't think they need and are not interested in.
It certainly seems likely there will many students who, in a chatbot life, learn to do whatever the bots tell them to do, and answer as the bots tell them - the right (bot) answer.
In Slovakia, they give oral test questions. Typically 2 or 3 hard complex topics from some 20-30 possible that the student is supposed to be ready for. Almost no cheating in the answering, tho there are cases of bribery, as well as "luck" for getting the easier rather than harder questions. There will likely be more of these.
Great ideas! One of the most exciting uses I’ve found for ChatGPT is just, as you suggest, getting into a dialog with it about interesting things, asking for bullet points, outlines, timelines, compare-and-contrast, sources, experts, etc. It’s a tool made exactly for the kind of active learning that aligns with effective learning.
I find this stuff so incredible, and I can’t imagine that the world we live in is going to be transformed by these tools. I’m trying to figure out why that isn’t the case, and how to put the brakes and be skeptical of my own excitment. But every time I interact with ChatGPT, Bing, Midjourney, etc. I’m just completely blown away.
This is worth repeating: “I think that we were already in a technological environment that favors autodidacts over traditional classroom learning. Chatbots take that to the next level.”
Education is a foundational component of future development and it seems as we a poised for exponential improvement in learning. Everyone will benefit from this.
Nevertheless, the best and the brightest will reap most of the rewards and that will certainly bring out the “equality thugs” to bemoan this advancement.
I note that you don't bring up the problem of chatbots being optimized to mimic human writing, not presenting a range of "true" information or accurate representations of it. Even prompting to answer as though it were a top economist is roughly as functional as pointing an undergrad to a wall of books by "top economists" and telling them to write an answer to your question. Neither one actually understands the answers and are just copy/pasting things that seem plausible. If the answer is internally contradictory, oh well. If it just makes stuff up to make it sound better, oh well.
Maybe an even better example is when kids ask their parents questions they don't know the answer to, and the parents just make something up that seems plausible. Terribly funny when Calvin's dad does it, but not so great if people are getting large amounts of information that way. Checking the sources is important, if you get them, but how many people are actually ever going to do that? Hell, peer reviewers never check the citations in academic papers it seems, so...
This would be fun but I think it accomplishes a different task than standard lectures do, and it also flattens critical thinking as most students will pick up the "standard" views and facts, and there really wouldn't be time for reflection and asking questions.
The teams will have to learn how to do this. For example, constructing prompts that focus the chatbot on reliable sources, e.g. by saying "You are leading economic historian" can prevent the chatbots from giving oddball responses. Double-checking sources is another approach, but I think that good prompt construction really takes care of it.
I'm skeptical. Airline pilots, for example, have to deal with automation complacency; as the autopilot runs more and more of the flight, the pilots become less and less aware of what's going on (particularly with fly-by-wire planes, where there's less feedback about what the autopilot is doing for you), and when the autopilot shuts down (because the plane is outside the control envelope), the pilots are thrown immediately into a crisis situation.
The same thing applies with self-driving cars; drivers are expected to remain fully alert and able to take over instantly.
But humans just don't do that. Which is why planes fall out of the sky and cars crash into pedestrians; and why students will mindlessly regurgitate what they've been presented by the computer.
Heck, head down to your local bureaucracy and try to get the clerks to do something that the computer doesn't have an entry for.
"I am optimistic that chatbots can prove to be the most important innovation in education in my lifetime. "
I'm surprised by this much optimism coming from you. Regardless, education is mostly about the student being interested. This will help some be more interested. Great. Why will it be significantly different in importance from Coursera,etc., traditional classes taught online, Babbel,etc., YouTubes, web searches, Wikipedia, etc.?
Why you think learning in 8 hours with a chatbot would be greater than many more hours over ~12 weeks using other methods?
And given your disdain for universities, you putting the chatbot in a classroom kind of blows my mind.
I will just repeat what I wrote on this substack a few months ago- the LLMs will be utilized effectively by the most talented people, and the rest of the people will use them as crutches allowing further atrophy of human capital.
"I could sit in front of students, give Chatbots prompts, and show everyone my screen. But that seems silly.
Better to have students do the prompting. "
Are you implying we don't need teachers anymore?
this is a fantastic idea. Got me thinking about ways to engage the people at my company with ChatGPT. I could imagine doing something similar as a training exercise that gets people familiar with the tech.
I like this idea. I think it would be good for students and teachers alike. Embracing AI in this way provides a good opportunity for students to learn how to best use the tech, lessons like the importance of asking the right questions and checking sources. And if teachers don't utilize these tools, they won't be able to take advantage of what AI has to offer or prevent outcomes they don't like, such as cheating on exams.
'Autodidact' is doing some heavy lifting here. Autodidact means self-teaching, but actually, everybody is self-teaching when they actually care about learning something. Practically everyone has gone on youtube to watch an instructional video on how to do something.
Autodidacts are more self-motivatatng than self teaching ("auto-proponits" maybe), and so will tend to figure out how to make best use of whatever tools become available to accomplish their learning goals.
But for most students (at any level of cognitive ability) it's motivation, not tools or techniques, that is the key issue. The problem is not lack of tools, access, or resources, which have been around for a long time in powerful forms and at low prices which would have astounded preceding generations. The problem is shaping situations and incentives to nudge the marginally un-motivated student to *want* to learn.
My hunch is that at this point we run into issues of hard-wired social psychology that make it difficult to substitute those social influences with computer tools. For example, it is instinctively easier for a member of 'the audience' to pay attention to some focal 'performance' when that member subconsciously picks up the subtle cues that the people around him are also paying close attention. It is easier to do what one imagines a teacher or professor or mentor wants (or what would impress them as an opportunity to show off) when one subconsciously calculates that affilitating with an authority figure or high-status individual would likely have personal benefits down the line.
Chatbots don't have status, can't do anything for you, don't make you cool or sexy when you tell people you use them. Interacting with a chatbot - in what is bound to be a typically solitary fashion - doesn't provide those social cues or incentives, and thus, doesn't help coach or motivate (at least, not yet, maybe soon!)
So, the autodidacts / autoproponits will indeed derive a lot of advantage from this new powerful tool, just like they did with all the previous powerful tools. But the other students will just use it like a search engine as an even faster way to get the answers to the homework or test or lines for the essay.
Perversely, I suspect that being even more able to outsource research, learning, and now even impressive composition to the chatbots will actually undermine the retention of actual skills or knowledge, just like people who were once able to use paper maps to navigate around a city found that they couldn't do it anymore after years of relying on smartphone GPS apps.
Surprisingly, I'm pretty sure most of this is not quite right. Yes, the self-teachers will use the tools the best, but I'm pretty sure many medium poor students will get turned on by the ability to get understandable answers to their questions, helping them more quickly understand the world.
Yes, many or most students will use it to get homework done as quickly (and with as little retained learning?) as they can. And even more will remember less - of the uninteresting stuff, usually what they have that they don't think they need and are not interested in.
It certainly seems likely there will many students who, in a chatbot life, learn to do whatever the bots tell them to do, and answer as the bots tell them - the right (bot) answer.
In Slovakia, they give oral test questions. Typically 2 or 3 hard complex topics from some 20-30 possible that the student is supposed to be ready for. Almost no cheating in the answering, tho there are cases of bribery, as well as "luck" for getting the easier rather than harder questions. There will likely be more of these.
This!
Great ideas! One of the most exciting uses I’ve found for ChatGPT is just, as you suggest, getting into a dialog with it about interesting things, asking for bullet points, outlines, timelines, compare-and-contrast, sources, experts, etc. It’s a tool made exactly for the kind of active learning that aligns with effective learning.
I find this stuff so incredible, and I can’t imagine that the world we live in is going to be transformed by these tools. I’m trying to figure out why that isn’t the case, and how to put the brakes and be skeptical of my own excitment. But every time I interact with ChatGPT, Bing, Midjourney, etc. I’m just completely blown away.
This is worth repeating: “I think that we were already in a technological environment that favors autodidacts over traditional classroom learning. Chatbots take that to the next level.”
Education is a foundational component of future development and it seems as we a poised for exponential improvement in learning. Everyone will benefit from this.
Nevertheless, the best and the brightest will reap most of the rewards and that will certainly bring out the “equality thugs” to bemoan this advancement.
This is maybe not quite as ideal as a "Young Lady's Primer" - but far far more realistic. Could even be done, technically now.
Using the hack-a-thon model is a fantastic, maybe paradigm changing idea.
I can easily imaging groups of home-schoolers trying this out.
Even have school teams, competing with other teams, and voting on which ones they think are the best at the end (ranked order 1-8).
I do think some "teaching how to use a chatBot" examples by the teacher would be good.
Ethan Mollick has written a number of posts on this topic. For example: https://oneusefulthing.substack.com/p/using-ai-to-make-teaching-easier . As have, I believe Tyler Cowen &or Alex Tabarrok.
I note that you don't bring up the problem of chatbots being optimized to mimic human writing, not presenting a range of "true" information or accurate representations of it. Even prompting to answer as though it were a top economist is roughly as functional as pointing an undergrad to a wall of books by "top economists" and telling them to write an answer to your question. Neither one actually understands the answers and are just copy/pasting things that seem plausible. If the answer is internally contradictory, oh well. If it just makes stuff up to make it sound better, oh well.
Maybe an even better example is when kids ask their parents questions they don't know the answer to, and the parents just make something up that seems plausible. Terribly funny when Calvin's dad does it, but not so great if people are getting large amounts of information that way. Checking the sources is important, if you get them, but how many people are actually ever going to do that? Hell, peer reviewers never check the citations in academic papers it seems, so...
This would be fun but I think it accomplishes a different task than standard lectures do, and it also flattens critical thinking as most students will pick up the "standard" views and facts, and there really wouldn't be time for reflection and asking questions.
Cool idea!
How will you avoid the hallucination problem?
The teams will have to learn how to do this. For example, constructing prompts that focus the chatbot on reliable sources, e.g. by saying "You are leading economic historian" can prevent the chatbots from giving oddball responses. Double-checking sources is another approach, but I think that good prompt construction really takes care of it.
I'm skeptical. Airline pilots, for example, have to deal with automation complacency; as the autopilot runs more and more of the flight, the pilots become less and less aware of what's going on (particularly with fly-by-wire planes, where there's less feedback about what the autopilot is doing for you), and when the autopilot shuts down (because the plane is outside the control envelope), the pilots are thrown immediately into a crisis situation.
http://www.vanityfair.com/business/2014/10/air-france-flight-447-crash
The same thing applies with self-driving cars; drivers are expected to remain fully alert and able to take over instantly.
But humans just don't do that. Which is why planes fall out of the sky and cars crash into pedestrians; and why students will mindlessly regurgitate what they've been presented by the computer.
Heck, head down to your local bureaucracy and try to get the clerks to do something that the computer doesn't have an entry for.
"I am optimistic that chatbots can prove to be the most important innovation in education in my lifetime. "
I'm surprised by this much optimism coming from you. Regardless, education is mostly about the student being interested. This will help some be more interested. Great. Why will it be significantly different in importance from Coursera,etc., traditional classes taught online, Babbel,etc., YouTubes, web searches, Wikipedia, etc.?
Why you think learning in 8 hours with a chatbot would be greater than many more hours over ~12 weeks using other methods?
And given your disdain for universities, you putting the chatbot in a classroom kind of blows my mind.
I will just repeat what I wrote on this substack a few months ago- the LLMs will be utilized effectively by the most talented people, and the rest of the people will use them as crutches allowing further atrophy of human capital.