I graduated college June 2026, so I have some insight into what’s going on in universities. dvd’s has a better POV because he gathered information more systematically and I self-selected CS majors AI-risk-concerned students, but I figure I can share some of what I saw.
I can confirm that #2, #4, and #8 are common among my classmates. I haven’t seen the others.
I’m surprised by #1. People have been using AI for more and more difficult things over time. You can’t productively have ChatGPT 3.5 help you with anything difficult, and until very recently AIs couldn’t consistently read text in images.
#5 is a good point. Unfortunately, not all AI leaders see the issues, and they all think that it would be best for them selves to win, even though they are putting everyone at risk. I can understand why students would believe #5.
I’m guilty of #9. Even though I’m young and clueless and lack a lot of skills because of AI assistance, I feel worried for those younger than me who have relied on AI for an even longer portion of their lives. Coding by hand 3 years ago and coding with light AI assistance 1 year ago taught me a lot, and I’m prouder of the things I made without AI.
I have hoping since 2024 that my university group, Northwestern University AI Safety and Governance Group, would find it easier to recruit as AI risks became more obvious, and I had hoped that younger students would have an easier time understanding the power of AI (if they are “AI-native”, which you suggest they are not). In the following two years, I have been confused when DeepSeek-R1, o3, Claude Code, and Mythos came out, and I saw no influx of interested new members. Thank you for casting light on what has been going on with the students I have been trying to recruit.
I’m surprised by #1. People have been using AI for more and more difficult things over time.
I think it is less surprising for current early university students (esp. outside top students/schools) because they probably weren’t feeding the AI of three years ago (GPT4) tasks that far beyond its capabilities. High-school-level homework, short essays, etc.
From their perspective it might be: “AI did a good job on my homework on 2023, it still does a good job on my homework in 2026.” They never tried the college-level homework problems in 2023, let alone were trying to use it for software engineering, real research tasks, etc.
Thanks for writing this!
I graduated college June 2026, so I have some insight into what’s going on in universities. dvd’s has a better POV because he gathered information more systematically and I self-selected CS majors AI-risk-concerned students, but I figure I can share some of what I saw.
I can confirm that #2, #4, and #8 are common among my classmates. I haven’t seen the others.
I’m surprised by #1. People have been using AI for more and more difficult things over time. You can’t productively have ChatGPT 3.5 help you with anything difficult, and until very recently AIs couldn’t consistently read text in images.
#5 is a good point. Unfortunately, not all AI leaders see the issues, and they all think that it would be best for them selves to win, even though they are putting everyone at risk. I can understand why students would believe #5.
I’m guilty of #9. Even though I’m young and clueless and lack a lot of skills because of AI assistance, I feel worried for those younger than me who have relied on AI for an even longer portion of their lives. Coding by hand 3 years ago and coding with light AI assistance 1 year ago taught me a lot, and I’m prouder of the things I made without AI.
I have hoping since 2024 that my university group, Northwestern University AI Safety and Governance Group, would find it easier to recruit as AI risks became more obvious, and I had hoped that younger students would have an easier time understanding the power of AI (if they are “AI-native”, which you suggest they are not). In the following two years, I have been confused when DeepSeek-R1, o3, Claude Code, and Mythos came out, and I saw no influx of interested new members. Thank you for casting light on what has been going on with the students I have been trying to recruit.
I think it is less surprising for current early university students (esp. outside top students/schools) because they probably weren’t feeding the AI of three years ago (GPT4) tasks that far beyond its capabilities. High-school-level homework, short essays, etc.
From their perspective it might be: “AI did a good job on my homework on 2023, it still does a good job on my homework in 2026.” They never tried the college-level homework problems in 2023, let alone were trying to use it for software engineering, real research tasks, etc.