A new dean has taken my place, and I have returned to the professoriate at Illinois Tech, ending thirteen years in administration, six as dean and seven as department chair at Georgia Tech. I won't rule out more administrative roles in the future, but only if the right role presents itself.
I'll teach intro theory in the fall, my first course since 2018, and take a sabbatical in the spring, mostly at Oxford. I plan to focus on writing, hoping to get out another book or books and other projects. It will be hard to go back to traditional computational complexity research, the field has changed considerably. I plan to spend some time understanding how AI changes the way we think about computation. Particularly why we see many of the benefits of P = NP while cryptography remains secure.
Also for the first time in 13 years I don't have a "boss". Technically I report to the department chair, who until a few days ago reported to me. But tenure protects my job, I choose my own research agenda, and teaching and service assignments are more of a negotiation than a top-down decision. Freedom!
For the blog, I have held back talking about the inner workings of universities while I had administrative roles. I'll now be more open in giving my thoughts, at least in general terms.
The next chapter begins...
Congratulations!
ReplyDeleteCongrats Lance!
ReplyDeleteThat's Oxford, England, right? (as opposed to some other Oxford). I look forward to seeing you! Greetings from Oxford
ReplyDeleteThanks for sharing! About Oxford and UK out of all places, I'm less confident, Paul will tell otherwise of course. I'd thought the US is still exciting but then again the exciting places you already know and have visited.
ReplyDeletePerhaps it would be interesting to nice to hear about how much complexity theory changed and the challenges of getting back into. Is it analogous to a field one opens up via a landmark paper and then diverted attention to different areas but after a decade reconsidering coming back?
" I plan to spend some time understanding how AI changes the way we think about computation." Well said. I hope this direction gives useful things.
ReplyDelete"while cryptography remains secure."
ReplyDeleteMy guess is that cryptography will remain secure for a long long while. For starters (wiki, I think, reports that) there were already two quantum-safe crypto techniques known in 1994, when Shor discovered (invented?, created?) Shor's Algorithm. And lots more have been invented since.
Two points on AI. In terms of language, LLM output most resembles the speech of patients with Wernike's Aphasia. That's a disease in which the parts of the brain that think seem to have become disconnected from the parts of the brain that turn thoughts into speech. It's called a "fluent aphasia", since such patients speak fluently and profusely, but without meaning. (This result, obvious as it should have been to we AI critics, was from work done at the University of Tokyo.)
Point two is that you should read, and try to understand, what Gary Marcus* has to say**. It's correct and he works hard to be less in-your-face than your Tokyo correspondent. For example, Marcus wouldn't say, though he should, that "LLM based AI is an inanely stupid parlor trick, and when the smoke clears, will be seen as the most god-awful waste of time and money in Computer Science history." Like I said: I'm too in-your-face.
*: https://garymarcus.substack.com/
**: In a recent article, where I would say "LLMs have no mechanism for relating text to reality", Marcus points out that LLMs do not have an internal model of the world, and therefore cannot evaluate the truth/falsity of the text they output.
Congrats!
ReplyDelete