How the time goes

My original idea when I started this was that I'd blog every morning when I got in to let you guys know what was going. And that worked --- while I was living at the KLI (that is, for about 3 weeks). Now I take the train in and I am nearly always working on the train, so when I get to work I don't shift back to non-work stuff. I think the 50 minutes a day I spend on the train account for a disproportionately large amount of what I get done, not least because they also help me get into my work in the morning without the distraction of email etc.

So an interesting thing I learned about science since I've been here is that there are really two very different steps to it 1) figuring out how things work and 2) convincing other people. Originally you think 1) is all that matters, but I had already realized there must be a second step with communication, or you were achieving nothing. Science is about societal progress in understanding. Individual enlightenment is not an end goal, just a part of the overall mechanism. Sometimes when I am feeling AI / futurist or memeticist I wonder whether individual enlightenment really matters at all -- so long as we are improving our predictions maybe we are doing science, even if we are doing that improvement through local improvements of theory without anyone really globally understanding what's going on, because a theory is too big for any one person's head.

Well, anyway, just sticking your idea on the web sort of communicates it, but you need to be more convincing than that to actually do good science. So a lot of what you spend your time
  • running unnecessary-for-discovery, over-the-top experiments to make your findings really, really obvious even to naysayers
  • producing attractive-looking and very clear figures
  • getting your work "certified" by publishing it in the best journal you can.
With respect to the figures, I am unashamed to say that I am getting a huge amount of help from Will, who has of course made data & methods his field of expertise. Nevertheless, I spent a lot of time over the last few months running new large data sets of simulations and hacking on the statistical code Will wrote for me (in R) to make the figures look even better & clearer. I finally finished this about two weeks ago, so I spent a couple of days on the week of the 26h completely rewriting a paper that had previously been rejected from a top journal to make it clearer, to use the new figures, and to update it with the research of other people that had been published in the last few months (this area is very hot!) Then I sent the paper to a few of my colleagues. One was Nick Priest, who was sure I should get it published & promised to look over the paper. This is the paper on the evolution of culture.

Then I switched to working on another paper. This one is just for a conference, though eventually it should become a journal article, and also (sooner) it will become a funding proposal. The conference deadline was Sunday night and I started working on it Wednesday, but fortunately I found I'd done some more work on it than I remembered, probably when I was on the plane sometime and should have been working on something else. Having laptops is great, you can always be writing down ideas. As it happened, I changed all my old text, but having even the sketch of ideas there did really accelerate the writing. This paper is about looking at the impact of aging on task learning, and what that tells us about how we learn and think. It's an extension of some work I originally did over Christmas in about 2000 when I was working on my MIT PhD and just wanted a problem to write an AI program around to test some of my new AI code. I wound up doing a lot of the "individual enlightenment" part of of that while I was in Harvard, understanding my own results and how they fit into the rest of the field but it didn't get into a good journal until 2007, partly because the first years of a new teaching job are always terrible & take loads of time, but also because I was still learning about communicating & convincing. Well, that paper is out and the work is done, and the new paper moves on to a new but neighbouring problem.

Now, there's a group at Prague (Charles University, one of the oldest in the world) founded by Cyril Brom that does more work on my PhD research (the AI action selection stuff) than any other group in the world does (that I know about!), including mine. So a week ago Friday they organized a whole day of talks about action selection & its applications, two of which were by me. So last Thursday I spent the morning working on email and the aging/learning paper, then Thursday afternoon I spent the 5-hour train ride to Prague writing the two talks I was going to give. This is despite the fact they were just modifications of talks I'd given before. Writing a talk is again another way of really thinking a problem through and how to communicate it better, and in this case I was bringing a few short talks together and trying to make a better and clearer overall talk than I had. This is part of why I love teaching (well, at least in my own area), because it makes you really understand things more deeply, and can help you move your own understanding along. And if you are at the front of your field, then moving your own understanding helps move the field too. That's one reason why it's so important that research and teaching both happen in universities, and researchers aren't isolated in laboratories. I am excited about one of the talks I gave them, the one I worked on most, called "Time for AI" -- it is about not only how to build better AI but why natural intelligence is the way it is, because computation takes time. I have already written a sort of a magazine article about that which has been submitted, but I think this may again turn into a grant and a bunch more articles.

The second talk I gave them was about the evolution of culture. I've been giving that talk for a while now & it has gotten fairly good, so that is why I was able to write the new version of my paper in two days. All these things feed off each other. Anyway, so I didn't have to rewrite that much.

Friday the students showed me their new work, which is great. Some of the students have a bit of a problem which is that they are leading the world at something -- providing educational tools for programming students based on building AI game characters -- but even though a bunch of smart people have been working on their project for years, and even though they are from the best university in their country, they can't believe they are actually the best in the world at something, so if they are they figure it must not be worth doing. This is something people at MIT never had a problem with, more the opposite --- they often assumed they were leading the world & didn't need to read other things. It's very frustrating in Prague, some very good people aren't sure they want to stay on in academia, which is fine but not if it's for the wrong reasons. But on the other hand, there are other people who are very enthusiastic and not at all worried, so I suppose those will be the people who will make the difference and carry the project forward. When I was talking at that AI & games meeting last month in Bradford I mentioned their work & several people went on at length saying that it was great work & they were excited about it, so I told the group that. And my talks went well, so hopefully that will help keep all the students excited. After my second talk just a couple people stayed and showed me some of their new more biological research they were doing, so that was cool too.

Saturday on the 5-hour train back of course I spent the whole time on the paper due Sunday. Will was in Vienna so we actually didn't do that much work that evening, but Sunday we met my PhD student & his partner for brunch, then went into the KLI and got a ton done the paper, then went to dinner with friends in the 9th district, then I was up until 1am finishing & submitting the paper. The last hour was just dealing with the submission software. Again, I had help proof-reading the paper, this time from my friend from Harvard now in Oxford, Mark Baxter. I didn't really have time to fix everything I should have, but since this is a draft conference paper I will have time still to get it right in the future. I actually had to do some of the analysis at the last minute and Will helped me again with the coding because I'd forgotten a language it had been so long since I used it. Science is a social process and it goes so much faster now, and you can collaborate with the best people you know who are interested, wherever they live. We must be building culture faster, much much faster, since we have the internet. Again, it's amazing to think about this.

Monday I still hadn't heard from Nick but I talked at length with my two other students, both by instant message. And finally Tuesday or so I heard I could go ahead with my paper. So then came a day, I think it was Wednesday, where I spent the entire day submitting the paper! (Well, except for some time on email as always.) Again, this must be such an improvement that allows the journals to go through so many more papers than they used to, but I had to spend a ton of time researching their own editorial board to suggest people they might want to use to edit the paper, and I had to get all my files into the right formats and figure out how to fill in all the forms. I think I worked about 6 solid hours just on the submission process (not counting lunch, other email, etc.)

So now I'm finally working on what I expected to work on all month (so I'm not too many days behind, I started Thursday) which is writing grants so I can have lots of students and postdocs when I go back to Bath. I want a large group, because even though I do OK with support from all over the world, students are much more affected by the people they see every day, and I think mine so far have been too isolated from people doing similar things. So if I get some postdocs in they will help set the tone & give advice etc. The reason I haven't before is because writing the grants took so long (esp. when so few can get funded) and all that time was time you weren't really doing research. But now I think I have got the hang of it better and also I have a better probability of success with the publications I've had recently and the fellowship too. But it is hard to stay focussed on writing the grants when there are so many exciting models to write and papers too. But hopefully I will be back to doing that by April.

I'm just hitting on the highlights and hinting at the fact there is other stuff you have to do to. It's an endless amount of email and requests. I have to review other people's papers (I am committed to 14 conference papers and two journal articles in the next couple months...) even just agreeing or rejecting these offers takes time, let alone doing them. And you have your direct infrastructure. One of my coworkers has had a lot of problems that I've seen other coworkers have before and so I have wound up spending time trying to help out, as are other people at the institute. And I have been given a course I have been asking to teach for years by Bath, so I will finally actually teach something I'm an expert in, but I have to write the full "unit proposal" and things like that.

The next couple months are going to be bad again in that Will & I won't see each other much at all. But in mid April we are going to take a proper "no laptops" week off for our 10th anniversary and go to Turkey. Although we've both been asked to give talks while we are there and probably will. But we'll email the talks! We are still trying hard to figure out how we can live together, but it isn't easy to arrange.
1