Want a good intro to AI? This is it.

One of the best lines in Marvel Cinematic University history, if I do say so myself, is this one:

Indeed, I often think about how “all words are made up” and therefore, they necessarily take on new meanings across times and cultures (except for “fetch” which no one ever made happen).

One downside to this is that words can be used so often, in so many ways, by so many people, that they start to lose meaning. I fear “AI” (artificial intelligence) is becoming like that, with everyone (mis-)using the word in myriad ways (see Emily Bender for great work on misuse and poor thinking about AI). All this confusion about AI is why I was pleased to come across Ezra Klein’s podcast with Demis Hassabis, chief executive of Google DeepMind. If you want a clear, thoughtful, and meaningful discussion of what “AI” used to be, is now, and could be in the future, this is it.

I’m on the record saying people are often worrying about the wrong things when it comes to AI and education. I think AI has tremendous potential, and we’ve used it a bit in our work, but I am skeptical it is going to completely replace teachers or make obsolete the need for people to…um…know things…themselves…like…without a computer. I suspect in the future AI will be a really important tool that students and teachers will need to know how to use, but it’s not replacing human thought.

What I’d like to see is work to move from AI-based instruction (e.g., Khanmigo) on closed-ended problems (e.g., solving a math problem) to more open-ended problems (e.g., how to curate accurate sources of information). When AI starts being able to teach people how to acquire open-ended problem-solving skills, that’s when I’ll get really excited. Also, I’d like to see AI in education trained not with “right or wrong” reinforcement but rather reinforcement for identifying a student’s current learning trajectory. More on that another time… :)