Every now and again I'm asked to be an outside member on a thesis committee. Every school I know of requires a member from outside the department and often outside the university. They're a lot of work, but they can be a lot of fun. At first you run into the imposter syndrome, but you can ask questions to clarity from the candidate as well as their advisor and even other people in their department. It's a neat window into an area you didn't know much about.
But you're there to bring your expertise and offer constructive criticism
About 20 years ago I was on doing this at a nearby Ivy. The candidate was excited by the work and very articulate, but as I came up to speed it was clear their statistical interpretation was just wrong. A few questions revealed they didn't know the basics and they should have had at least a year of statistics and a good probability course. Not knowing how to drive, they used a commercial statistics package. (the graphs were very pretty) I told them they'd have to do it right or not have it in the thesis. This turned out to be a problem. The defense was in three weeks. Their advisor agreed and the candidate spent the next four or five months learning statistics and reworking the thesis. The final defense was flawless, but I could tell they weren't happy having lost all that time, Then, a few years ago, I got a call out of the blue. It was the candidate, now a professor at UCSD. He thanked me for holding his nose to the grindstone - "it was one of the best lessons I had in college.." You don't get calls like that often.
In the last few weeks I've spent time with some other folks interviewing professors and students to get a sense of modifying the school's honor code as well as general instructions to students and faculty. It's clear this is an on-going effort as people learn and technology changes, but the experience reminds me of being an outside examiner. It's also clear the purpose of education at this school is to inspire students to follow a path of mastery rather than achievement. It's pretty clear that a lot of AI (I use the term loosely as it's mostly a marketing term at this point) is being used without much thought of costs vs rewards. Sometimes these tools are appropriate and sometimes they're wrong. Here's an analogy I proposed:
Almost everyone has a microwave oven now. Most of us use them to cook frozen meals, reheat leftovers, boil water, pop popcorn etc and some interest uses like cooking corn on the cob have emerged. They’re extremely useful in these domains. When they were less common - the period from the 60s through the mid 80s - manufacturers pitched them as general cooking tools that save time. Cook pot roasts, chickens, pies, etc.. there were special tools like temperature probes and browning trays, but the physics just doesn’t work and people generally gave up after a few attempts. Currently AI, in particular GAI, is like the early microwave ovens. It will take time for people to discover the best tasks for it. We’re still at the cooks your whole Thanksgiving dinner - or at least it will next year stage.
Without going into my personal experiences I offer an example of a serendipitous learning when a wrong tool turns out to work. At least it was serendipitous to me.. the person who pulled it off knew exactly what she was doing. First find a recording of the Preludio of Bach's E Major Partita for Violin. It's very famous so there are any number of good recordings - I like Isabelle Faust's interpretation, but your mileage may vary. After you've listened, try this version by Tobie Miller. She takes some liberties, but you can tell who wrote it and her playing is lovely.
Comments