Are Students AIs?

We end this year (2022) with an explosion in AI of many types. AI has been truly democratized and the lay public now interacts with it through sites such as ChatGPT and CharacterAI. 

I have been teaching for about half my life and it seems like being a professor is like being an AI. (1) We spend many years getting a PhD (DL: deep learning). (2) We teach students and engage them in deep learning, while having them learn from their mistakes. We also learn to teach by exploration and exploitation (RL: reinforcement learning). (3) We train students to organize knowledge (unsupervised learning). (4) They learn from a many examples (supervised learning) but also from brief experience (few-shot/zero-shot learning). (5) We engage students in knowledge production and creative work (generative AI). (6) We train students to learn how to learn and adapt to changing conditions (active learning). (7) And, we teach students to provide reasonable explanations for their work and actions, be honest and fair, and be ethical (responsible AI). 

In many ways we find that machines outperform students at many tasks but not as a holistic multipurpose intelligence, though multi-task agents already perform extremely well. And both display some common sense but not as much as we would like, though in all fairness, humans as a whole are hard-pressed to exercise common sense (sound judgment in practical matters). But most certainly, our students are better than machines at reasoning ("the action of thinking about something in a logical, sensible way"). but much worse than machines at memorization.

It is strange to be mapping AI concepts to teaching and learning, when in fact, the AI training nomenclature has arisen from human learning methods. But it also hides the sad fact that today we spend much more time on how we train AIs and machines to learn than on how we teach children. When we look back at 2022, we congratulate ourselves on how far AI has come, but the most salient note about school education is a lament about learning loss since the pandemic. Our machines have evolved but our students have regressed. Optimism for AI in 2023 is everywhere, but there is only pessimism for our education systems. We make huge investments in the former where the human capital is very well paid and we make none in the latter, as teachers are paid progressively worse in real terms. 

We need to invest in training the best intelligences we know, that run the best hardware there is. Maybe we can use AIs for that? I hope 2023 can be different.