Transcription
In an AI future, does Math 55 could continue to be the course that it is? I mean, what I I guess what I want to know from you is how does a great institution like Harvard navigate a world in which knowledge is going to be a commodity? What what what is an education a Harvard education going to look like when I don't know for all I know some AI model could do Math 55 in like five minutes?
Yes. Well, it it's a great question and I think you phrased it exactly the right way and the answer has to be of the form of what is it that's not a commodity in the future. What is a unique skill? What is a skill you need that AI itself can't fully satisfy? And I'd ask all of you to think about the answer to that and think about how you viewed generative AI say a year ago. Now around a year ago maybe 18 months I would have said what we need to teach our students is prompt engineering that is how to ask good questions of the model. Then the models became so good that they could help steer you into crafting a good prompt and finding out what you wrote. So all of a sudden that's not really a skill and uh you're all reading a lot about AI and I I would just say a few things about how I look at it and I think my colleagues do.
Number one, the capabilities of the models are changing so rapidly to say that there is some particular skill that's going to be evergreen that's going to matter forever is pretty hazardous because these capabilities change so much and especially with agentic AI where there are these big increases in the complexity of the kinds of tasks that AI can uh carry out. But if I and and I also believe that one implication for universities is we need to prepare the people who learn with us for a life and a career where their skills are going to need to evolve. They'll need to learn new things over time. And I think more and more we've talked a lot about over the years about lifelong learning. It really means something different today. And maybe what we should be thinking about is what our contract is with our students. Should we make sure that they have access to a lifetime of opportunities to reskill and to figure out exactly how they can best spend their time, not all of it work.
I can't resist saying that there's one image that always comes to mind for me and I don't know how many Pixar movie fans there are here but I remember in Wall-E where the most human characters were robots and the humans were in these lounge chairs basically having every need taken care of for them. That's not a future we want and we need to make sure that we can do better. But there are certain things that I think are going to be really important and that is your ability to detect the truth.
And for many people a disturbing moment recently was when OpenAI publicly contemplating contemplated incorporating advertising into their models. Now, I don't have a real objection to advertising per se, but you think about the conflicts that that sets up and how will you know whether the answer that it gives you and these models have personalities and they do know how to please? How to know when you're being manipulated? How to know when the facts that the model is bringing forth are really true or at least a reasonably comprehensive uh representation of the truth. You all need to have skills to do that and we do have ways to teach that and I think that is going to be incumbent on all of us.