I would say that on the whole, Chat-based LLMs ("AI") have been a large negative impact on higher education. They are good enough now that mediocre to poor undergraduate students (or whole institutions with average students) will receive better grades on assignments generated by AI than if they did it themselves (at least in the disciplines I am familiar with). Or, if not better then comparable with no effort expended.
The short-sighted benefits are just too overwhelming that even for students who arguably have skin in the education game (most do not and merely are there for credentialing) , it is a difficult battle - akin to the generation(s) who recognize that the type and amount of their social media use is problematic but react with resignation.
There is a difference between 'knowing' and 'doing things' - the consequences of AI are different across these.
Some other thoughts:
- I do not think there are any viable solutions to the AI 'problem' in education given the current structure of higher education.
- The affordances, the motivations and practical, if not overt, purpose of 'education' cannot meet AI head on.
- A stronger more explicit distinction needs to be (re-)drawn between what used to be 'vocational' and 'liberal' or 'higher' education. And a narrowing of the purpose of higher education (which would, in itself, be very disruptive)
- There needs to be an explicit addressing of AI in both curricula -- in terms of both practical training on its use as well as the pitfalls and downsides-- in terms of self-interest (e.g. if you have machines do the work to set you 'free' then you run the risk of becoming slaves to the people who make the machines (paraphrased Dune).
- Complete change in approach and most importantly assessment -- I am not sure this can be done in a system where grades are still the litmus test for learning.
- This educational crisis has to be addresssed BEFORE college- otherwise it is probably too late.