First, if one student is falsely accused of cheating while 10 are busted who actually cheated, that still doesn't justify anything. Innocent people being punished for things they're not the least bit guilty of is sort of a founding principle of our society.
Second, educators need to adapt to this new reality. I look forward to the follow up, and I've wanted to write about that for a long time now. I have friends who have been in (or are currently in) academic administration roles, and professors, and so on. I want to let them do most of the talking.
And, I hate to be gauche (by sharing my own article), but Sal Khan is also really onto something, although it's a pretty big can of worms:
Another great article, Roberto. Looking forward to the next one about how teachers are going to handle AI generated content. The coming AI-saturated academic year is hurtling down the tracks at teachers like a freight train and the education system is just. not. prepared.
AI does present a pivotal opportunity for educators to rethink outdated teaching methods and reinvent learning for the modern era. Rather than viewing AI as a threat, teachers can embrace it as a catalyst.
Transitioning from pure content-providers to mentors guiding students to think critically, work creatively, communicate clearly, collaborate meaningfully and contribute positively to the world. Exciting times ahead!
Agreed Pascal. This is an important point in the second part. We'll go through this sooner or later, better doing it with a mindset to take the most out of it. For now, I think only the most pioneering teachers will take this as an opportunity.
Most definitely! As a long-time technical translator (Japanese to English), my income was progressively affected by machine translation over the years. I adapted by using the tools to increase my efficiency. Teachers can do the same with the generative tools as they do with others, thus concentrating more on personal attention to individuals in class.
Loved it. One thing that I found inconsistent in this whole matter, the same kind of people who are claiming LLMs do best the Turing test already also claim they can build perfect detectors. You can't have your cake and eat it too. In practice, though, both are highly exaggerated claims. However I do believe we are much farther ahead in the first direction than in the second one, and the gap will keep widening.
Yes, although I wouldn't say there's incoherence there! The Turing test is about deceiving humans, while AI detectors are algorithms. An algorithm can detect something humans fail to.
I think you can say there's two ways of passing the Turing test: the weak and the strong claims. So if you say you can pass the Turing test weakly, that means in the sense that there is a high chance that you can trick a random human judge. So that doesn't necessarily mean that the AI is a strictly indistinguishable from human. It just means that for random human judges, there is a high probability that you will trick them.
And then you have a strong sense which means that actually the text generated by the AI is in principle completely indistinguishable from human text. And in that sense, you cannot have AI detectors that that work on that kind of output because by definition it is indistinguishable from human text.
I would suggest inviting a student to elaborate on the subject they wrote. Eventually requesting a permit to record the discussion, and then it will become rather easy to clear if they use the same language style, formulating similarly
As a retired ESL teacher who is actively engaged in the movement toward genuine AI (artificial persons), I believe the solution is "embrace and control" the generative tech, which should be both easy and inexpensive, and open up collaboration with tech developers rather than adversarial relations. Curricula should include instruction on use of the "AI" stuff and student use can be limited to an official in-house interface to the tech.
Clearly a new paradigm for pragmatic learning is in order. Rather than fighting against the machine, how about being assessed as a human-machine symbiotic team. This is all that any cutting edge human chess or Go player can aspire to. So essays are out. Symbiotic assessments of team play are in. This way “cheating” doesn’t even enter into it. The challenge for the human students are what they bring to the table to supplement the team effort. If it’s nothing they fail. So they have to be taught how to bring something. I’m pretty sure that’s what they’d have to do at Star Fleet Academy. So Yes, you can use a calculator.
One solution would be for educators to simply face the fact that AI writing is going to replace human writing, in the same way that keyboard generated text has replaced cursive writing.
When I was in college in the early 70s my roommate was taking a lot of math classes and there was some controversy about whether using calculators was cheating. That moment passed, and now it would seem absurd to have to do all your calculations with pencil and paper. What's the point of pencil and paper when anyone can buy a calculator for $10, or already has one on their laptop?
The writers I've seen across Substack seem to be engaged in wishful thinking fantasies that their human writing is special and unique and can't be replaced by AI. They don't seem to get that AI will be able to simply scrape their article and reword it, at a tiny fraction of what it cost them to write the article. A Substack blog that took years for a human to manually build will be copied and reworded in seconds. The entire Substack network will be duplicated, reworded, and republished elsewhere in a day.
Here's more fun news. Readers won't really care. As example, here on Substack less than half of the newsletters being mailed are even opened, let alone read. Those that get "read" are probably mostly being power scrolled. The average reader on the Internet is not going to care about the difference between AI and human generated content. Unless you the writer are the reader's mother, child or lover, they just aren't going to care.
I'm not able to say how soon this transition will be complete. Roberto would be much better qualified to comment on that. But unless we think there is some insurmountable barrier which AI text generators will never be able to leap over....
Human writing is dead.
If true, then college professors should probably start adapting to the new reality their students will spend their lives in.
I’ve had some truly abominable teachers. The worst were going through difficult life issues like a divorce or substance abuse. I think for parents it would be invaluable to have a daily summary of the teachers mood and consistency. AI also provides tools for the best teachers to reach more students. I think having a positive, nurturing presence in each students life that can act as a sort of educational body cam would help a lot of kids.
Nailed it.
First, if one student is falsely accused of cheating while 10 are busted who actually cheated, that still doesn't justify anything. Innocent people being punished for things they're not the least bit guilty of is sort of a founding principle of our society.
Second, educators need to adapt to this new reality. I look forward to the follow up, and I've wanted to write about that for a long time now. I have friends who have been in (or are currently in) academic administration roles, and professors, and so on. I want to let them do most of the talking.
And, I hate to be gauche (by sharing my own article), but Sal Khan is also really onto something, although it's a pretty big can of worms:
https://goatfury.substack.com/p/an-ai-tutor-in-every-childs-hand
"Innocent people being punished for things they're not the least bit guilty of is sort of a founding principle of our society."
100% agree
I wish everyone understood why this matters so much.
Another great article, Roberto. Looking forward to the next one about how teachers are going to handle AI generated content. The coming AI-saturated academic year is hurtling down the tracks at teachers like a freight train and the education system is just. not. prepared.
AI does present a pivotal opportunity for educators to rethink outdated teaching methods and reinvent learning for the modern era. Rather than viewing AI as a threat, teachers can embrace it as a catalyst.
Transitioning from pure content-providers to mentors guiding students to think critically, work creatively, communicate clearly, collaborate meaningfully and contribute positively to the world. Exciting times ahead!
Agreed Pascal. This is an important point in the second part. We'll go through this sooner or later, better doing it with a mindset to take the most out of it. For now, I think only the most pioneering teachers will take this as an opportunity.
Most definitely! As a long-time technical translator (Japanese to English), my income was progressively affected by machine translation over the years. I adapted by using the tools to increase my efficiency. Teachers can do the same with the generative tools as they do with others, thus concentrating more on personal attention to individuals in class.
Loved it. One thing that I found inconsistent in this whole matter, the same kind of people who are claiming LLMs do best the Turing test already also claim they can build perfect detectors. You can't have your cake and eat it too. In practice, though, both are highly exaggerated claims. However I do believe we are much farther ahead in the first direction than in the second one, and the gap will keep widening.
Yes, although I wouldn't say there's incoherence there! The Turing test is about deceiving humans, while AI detectors are algorithms. An algorithm can detect something humans fail to.
Yeah, I guess you're right.
I think you can say there's two ways of passing the Turing test: the weak and the strong claims. So if you say you can pass the Turing test weakly, that means in the sense that there is a high chance that you can trick a random human judge. So that doesn't necessarily mean that the AI is a strictly indistinguishable from human. It just means that for random human judges, there is a high probability that you will trick them.
And then you have a strong sense which means that actually the text generated by the AI is in principle completely indistinguishable from human text. And in that sense, you cannot have AI detectors that that work on that kind of output because by definition it is indistinguishable from human text.
So yeah, I agree with you.
An interesting subject for sure.
I would suggest inviting a student to elaborate on the subject they wrote. Eventually requesting a permit to record the discussion, and then it will become rather easy to clear if they use the same language style, formulating similarly
As a retired ESL teacher who is actively engaged in the movement toward genuine AI (artificial persons), I believe the solution is "embrace and control" the generative tech, which should be both easy and inexpensive, and open up collaboration with tech developers rather than adversarial relations. Curricula should include instruction on use of the "AI" stuff and student use can be limited to an official in-house interface to the tech.
I was going to write "Nailed It", but @Andrew Smith already did. ;-)
Idea - educators need an AI tool that smartly conducts oral exams and quizzes.
Clearly a new paradigm for pragmatic learning is in order. Rather than fighting against the machine, how about being assessed as a human-machine symbiotic team. This is all that any cutting edge human chess or Go player can aspire to. So essays are out. Symbiotic assessments of team play are in. This way “cheating” doesn’t even enter into it. The challenge for the human students are what they bring to the table to supplement the team effort. If it’s nothing they fail. So they have to be taught how to bring something. I’m pretty sure that’s what they’d have to do at Star Fleet Academy. So Yes, you can use a calculator.
One solution would be for educators to simply face the fact that AI writing is going to replace human writing, in the same way that keyboard generated text has replaced cursive writing.
When I was in college in the early 70s my roommate was taking a lot of math classes and there was some controversy about whether using calculators was cheating. That moment passed, and now it would seem absurd to have to do all your calculations with pencil and paper. What's the point of pencil and paper when anyone can buy a calculator for $10, or already has one on their laptop?
The writers I've seen across Substack seem to be engaged in wishful thinking fantasies that their human writing is special and unique and can't be replaced by AI. They don't seem to get that AI will be able to simply scrape their article and reword it, at a tiny fraction of what it cost them to write the article. A Substack blog that took years for a human to manually build will be copied and reworded in seconds. The entire Substack network will be duplicated, reworded, and republished elsewhere in a day.
Here's more fun news. Readers won't really care. As example, here on Substack less than half of the newsletters being mailed are even opened, let alone read. Those that get "read" are probably mostly being power scrolled. The average reader on the Internet is not going to care about the difference between AI and human generated content. Unless you the writer are the reader's mother, child or lover, they just aren't going to care.
I'm not able to say how soon this transition will be complete. Roberto would be much better qualified to comment on that. But unless we think there is some insurmountable barrier which AI text generators will never be able to leap over....
Human writing is dead.
If true, then college professors should probably start adapting to the new reality their students will spend their lives in.
I’ve had some truly abominable teachers. The worst were going through difficult life issues like a divorce or substance abuse. I think for parents it would be invaluable to have a daily summary of the teachers mood and consistency. AI also provides tools for the best teachers to reach more students. I think having a positive, nurturing presence in each students life that can act as a sort of educational body cam would help a lot of kids.