[Bostonian]

The elephant in the room is that many students are using ChatGPT to write their papers. Since they will use similar tools at work to create presentations, I'm wonder how much colleges should try to police use of ChatGPT by students.

[i_am_proteus]

In a class of any substantial size, it's very obvious who is using ChatGPT and who is thinking for themselves.

A large fraction (often 70%) of responses will be identical in structure and content, if not verbatim copies of each other. These are the people who have used ChatGPT. These are the people who tend to do very poorly on closed-laptop examinations.

Students have a choice of how much of their critical thinking they choose to offload to the computer, and how much they develop their own critical thinking skills.

Robust assessments in classes still paint an accurate picture of students' capabilities. Designing those assessments requires work for instructors. Pour one out for the lazy.

I suspect that in the coming years, industries will reward those who are capable of adding value beyond naïve parroting of LLM output. Pour one out for the lazy.

[flir]

> who is using ChatGPT and who is thinking for themselves.

You can do both. My son's using the chatbots as idea/synonym/high-level structure generators and general tidier-uppers. I'd be surprised if he's taking more than 10% of their suggestions verbatim, but it's great for rubber ducking.

[throwaway48476]

I've found chatGPT to be great for writers block as it's very good at giving me examples of what not to write.

[NemoNobody]

There will be a few that can write better prompts than others - based on your confidence I'm quite certain professors all across the world have already highly graded AI papers, while worrying about that very problem coming and failing others for it.

[aaplok]

My experience has been that performance has gone down significantly since students have access to genAI. I probably get a fifth of the number of high performers that I was getting before. So I'm not worried about the hypothetical tiny minority who can write good prompts. I am worried about the larger group who could have done good work but ended relying on chatGPT too much.

[i_am_proteus]

The real question is whether to meaningfully "grade" papers that can be written with the aid of a computer at all.

Alternative systems (which have existed before LLMs) include at-home work being required to earn a seat at the exam, but marked for feedback only, with the exam being the assessment graded for credit.

[yousif_123123]

At the same time using chatgpt or any LLM productively, and getting the most out of it to help you think and reason better is a skill on its own that requires development.

[NemoNobody]

You mean like teaching how to get Google to solve my math problems for me, instead of teaching me the logic behind the problems - what use do I have for theoretical calculus?

The only time in my life I've been unable to complete something mathematically, I was trying to compute accumulation of ROI for a shared money pool for crypto trading - I'm not even saying that correctly.

I couldn't enter the formula for the math I needed I to Google as I did not kno how to, have not owned a TI- anything calculator since highschool and completely forgot even the name for the genre of math that I was doing.

I kno that Google could do exactly what I needed tho and that was frustrating having to look literally every step of it all up

I think AI could have eliminated all of this.

[yousif_123123]

Yes it could be like learning how to use google to get information, or to learn the logic to solve something. I think the main under hyped aspect of AI currently is how useful it can be integrated with whole bunch of regular programming code and UI/UX to implement parts of functionality that can't be programmed explicitly. Like even if gpt4 is the last AI model, if it keeps getting cheaper and faster, the amount of cool features that can be with with some code and a few API calls in parallel or in sequence will add a lot of nice features to all the apps people interact with everyday.

Like my Dad uses multiple apps to get the information he wants for all the sports games that happen during the week, and no app is perfect. If he could prompt something to make a UI with the relevant info he wants displayed like he wants, he would be so much happier. It's little things like that that will make software way more customized and useful for people.

I think there's a lot of hype on AGI and they'll be more, but just as another tool in the programmer's toolbox, LLMs will allow better product to be developed (if people learn how to prompt and think with LLMs in mind vs dismissing all of it when they see a mistake)..

[monkaiju]

Citation needed

[eganist]

Wouldn't surprise me if schools go the route of forcing a cloud toolset (e.g Office 365 or Google Workspace) and measuring proof of work - in a very literal sense, not a cryptographic sense.

Ripe for being gamed but is probably the only way to do PoW without becoming wildly intrusive with personal devices.

[rahimnathwani]

  measuring proof of work

This is one of the ideas Sal Khan mentions in his recent book: interaction history with Khanmigo.

But there's nothing to stop you from using another AI system to interact with Khanmigo.

[eviks]

Or you could switch to proof of knowledge and use that old boring tech of a conversation. Also harder to game at such a low latency.