I have seen some articles, which talk about how AI comes to the classroom. We see how Chat GPT produces good quality documents including bibliographic sources.
In other cases, using any statistical mathematical software, the student no longer requires knowing how to solve them, and sometimes I do not even know what to do, because the "system" already proposes a solution. So, the teachers are prepared for what is coming?, To worry about the new teaching?, I already started making changes. The questions in an exam, changed the best procedure, to "explain why AI, proposed that solution?.
Carlos Jimenez-Gallardo I think tools like CGPT (and for coding CoPilot) help in producing initial drafts, approximate solutions, that otherwise take time but no specific curated level of skill. I recall seeing a similar discussion a few weeks ago related to an academic writing course, where the teachers were happy with CGPT because it let students overcome the first draft, then let them focus on iteratively improving/revising/re-editing, where arguably more skill is involved.
The generated solutions still need a lot of expertise to verify, e.g. in generated writing there are logical fallacies, sometimes utter nonsense, circular reasoning etc, so it's no assurance for a semantically valuable contribution. It just saves students time in coming up with a basic framework to improve upon, so education shifts to quality over quantity.
For code/problem solving, CoPilot / CGPT actually saves me time in figuring out how to call APIs, predict boiler plate code, that requires no creativity, but would otherwise cost me more time. Use cases where the problem is non-trivial, for those I'm on my own, it doesn't even come close to having a clue.
The case of students who do not scrutinize the generated solutions, well, that was already the case when they copied from Stackoverflow without thinking if the answer was valid, I recall software engineering publications mentioning 'Stackoverflow considered harmful' to refer to this aspect of reuse. As a counterexample, I learned a lot from Mathematica's step-by-step derivations for integrals when I got stuck, iow the situation where a student uses a tool without (initially) understanding how the tools works or comes up with a derived answer, is not necessarily bad. It is true that it is then up to me, at the end, to discover that knowledge, or at least verify the veracity of the answer.
In my alma mater courses (pre CGPT (~ 2012-15)) already had a heavy focus on inducing creativity, giving open research problems as questions on exams etc, not to test accuracy, but test research potential, problem solving, grasp/understanding, make associations, while penalizing memorization. In my opinion, this is more valuable as training/evaluation for courses whose aim is to improve and produce researchers.
However, I do understand that it is non-trivial for teachers/faculty to design courses like this, and probably not feasible for courses with > 100 students.
AI tools are still like kids. They're just learning and they are not always fully correct. Artificial Intelligence is good but not still that best to know how to train students. Human brain and emotions are still so much unknown for artificial intelligent tool. AI are still kid about that sector but you can consider them as a smart calculator which can solve big solutions.
@Naeem @Ben, despite being at the beginning, do you think they will be able to teach with this new "tool" in these new conditions, when they have been doing the same for years?
In some ways, the problem is not all that new. We had the advent of calculators, then the internet, smartphones and virtually instantaneous access to knowledge, AI is the next step. The essential question is: "What is the purpose of education?". The second in turn is: "How to verify whether the goal has been achieved?". If we want to increase the level of awareness and understanding of the world, then tools such as smartphones and AI are still just tools and their proper use should be taught. Verification of education progress should be based on direct interpersonal interactions/conversations, not automated in the process of tests/essays written at home, then the impact of tools is negligible.
- Badges
- Science topic