A few notes on Patrick Winston

I just found out (from a 5 day old tweet by Rod Brooks) that Patrick Winston died. This is a great portrait of him from The Tech. I'm astounded I will never talk to him again. The most recent time I did was in his office in CSAIL maybe 10 years ago. I was visiting just because I was visiting, and he was relieved since if I'd been an American academic visiting at that juncture after my PhD, I would have been asking for a tenure letter. But he asked me about my research and how things were going, and I could see on his face that I was doing OK, and that made me realise I was going to make it as an academic.

Just about five weeks ago I was talking to a young, brilliant, but not yet permanently-affiliated academic. I didn't really know her, but as she expressed her insights and concerns about how AI ethics money was flowing in and around Cambridge MA, I eventually felt enough empathy and trust to tell her the real secret – go ask Patrick Winston. He'll know what's really going on, and who to trust. I guess I made that suggestion too late.

A few other anecdotes inspired by that Tech obituary–the rest of this blogpost will make more sense if you take time to read that link, or else did AI at MIT.  I TAed 6.034 (Undergraduate AI, a third year course) in the recitation of a young prof whose wife was expecting their first child.  At MIT at least in computer science, such courses were taught with two enormous lectures per week to hundreds of students by the main professor (here, Patrick), then with two recitations also lead by full professors to groups of 30 or so students which were more interactive, and where the real learning was expected to occur. (The teaching load of an MIT prof at least when I was a student in the 1990s was leading one such giant course and one or two such recitations per year – each recitation also had a TA who did the actual marking.)  I wound up covering the young prof's recitations at short notice a couple times as they rushed to the hospital on "false alarms." I'd already taken ten courses on AI at Edinburgh and also the graduate level MIT course from David McAllester so I didn't have any trouble with the material in general, but Patrick's lectures were so wide ranging I had no idea what to hone in on for the undergraduates in those crucial one hour slots. So I went to the recitation of one of the professors I most respected at MIT anyway–Tomás Lozano-Pérez–who also happened to have his recitation directly before mine, and created an outline for my own tutorial from notes on his.

After the second missed recitation, Patrick became concerned about the students and checked up on me during the weekly course staff meeting about how I was handling the recitations. I said "I'm just going to TLP's tutorial and saying whatever he does." Patrick relaxed and laughed and said he himself and perhaps all the other recitation leaders should do the same.

If any of my own students, particularly who took programming or Intelligent Control & Cognitive Systems from me, read that Tech obituary I linked above, they will probably think they know where I got my teaching style from. They may be righter than I'd thought of, but I would have said I learned to teach mostly from Gerry Sussman, then Rod Brooks and Lynn Stein.  But Patrick always not only demonstrated good practice, but drew attention to the fact he was doing so. He was always aware whether it was an undergraduate lecture or a research seminar or an administrative duty or an exam-marking "party" (we marked all those hundreds of exams in one day at MIT, recitation instructors TAs and heads of course all together for as long as it took) that it was a teaching opportunity, and that teaching the content wasn't probably as important as teaching the process.

One day in the weekly 6.034 staff meeting PHW said "This week I teach this learning shit" &
I was surprised (not much of 6.034 is ML) "What learning shit?" "This arch shit." Having been taught AI
at Edinburgh I realised he meant his own PhD. Me (now excited to hear the lecture) "Oh! Your shit!"
Him (less excited): "yeah, my shit." Click the link for that lecture (a few years later.)

I need to get back to work, so I'll finish with just one other memory – the time I came in to work on a Saturday or Sunday morning and heard a weird noise and came out to find Patrick in a checked hunting shirt unscrewing screws from the wall of the halls of the eighth floor AI lab with a power drill. I think Rod Brooks was by then the head of the AI lab, any way I'm pretty sure it was Rod who had taken down all the framed pictures of famous old AI books written by people formerly in the lab, in the hopes someone would put something new up, but nobody did. It was now some months later, and Patrick came in from his home some considerable ways away (New Hampshire?) just to take the screws out of the walls (that the pictures had been hanging on) when he thought no one else would be there. I joked something like "are you sure you're qualified to do that?" It was the one time he looked at me like I'd disappointed him.