Thursday, January 17, 2019

The end of an era

One week from today will be the official end of an era for me. After nearly 23 years, more than half of my life, I will no longer have an account with MIT.

My MIT email address was my first "real" email address, at least in the sense that I can no longer remember definitely just what my high school email addresses actually were: you'd have to dig into the prehistory of AOL or old shareware repositories to find that information. I signed onto Project Athena in August of 1996, was indoctrinated into the joys of "mh", and began my tumultuous undergraduate career.

I stayed at MIT for 12 years as a full-time member of its community: four years as an undergraduate, one as a Masters student, one in a strange superposed state of both Masters and Ph.D. programs, which mightily confused the registrar's systems, five more years of purely Ph.D., and then a year of transitional postdoc while I figured out what to do, ultimately departing for my current employer of BBN in 2008.

Self-portrait as a postdoc, in my old office in Project MAC at MIT CSAIL
Even after leaving, however, I maintained my affiliation and strong collaborations. In the first couple of years, I was actually also still running the same MIT projects that I had been running while I was still actually employed there, and so of course I was always on campus to meet the students who were working for me. Other collaborations started up thereafter, and one way or other, it tended to be the case that I was working on campus at MIT at least half a day in every week. Keeping research affiliate status for me made a lot of sense for all involved.

Back in those early post-departure days, I used to still be much more involved as an alum in student group activities as well.  One of my long-running joys, which I still miss, was running live action role playing games with the MIT Assassin's Guild. More regularly, however, I also continued to be a volunteer librarian with the MIT Science Fiction Society, and every week would spend two hours as the on-duty librarian holding open the worlds largest publicly browsable collection of science fiction. I needed my card and my affiliation to be effective at those duties as well, and I enjoyed them much: the Assassin's Guild as a heated activity of creative passion and adrenaline, MITSFS as a cool oasis of two calm hours of mostly only reading.

When I followed my wife to Iowa in 2013, however, the actual "showing up on campus" part stopped happening. I resigned as a librarian, and I'd already mostly stopped writing and playing games, as new parenthood and professional travel began to squeeze that time more and more.  My collaborations have continued, but with me no longer actively on campus or needing special access to resources, there's not as much point in having me still maintain an active affiliation.

Sometime last summer, my affiliation failed to renew, and I didn't notice. When I got my account deactivation warning, I pinged the collaborator who'd been sponsoring me, but neither of us got around to following up. And really, the fact that it just wasn't making my triage list as "important" any more was the sign that it was time to let go.  I'm no longer an active alum, and I don't need a research affiliate status to be an effective remote collaborator, after all.

And so, over the past two weeks I've been packing up to go electronically. I've redirected my non-BBN mirror of my professional webpage away from MIT and over to GitHub. I've copied over all of the material from my old Athena account (finding and revisiting some amazing old memorabilia in the process). I've even gone through every email received at the old account in the last year and switched over all the ones I cared about. I'm as ready as I can be to let go.

Goodbye old friend, old email address. I never like to truly let anything be gone, but I'm not there any more, and at least I've still got my alum account.

Saturday, January 12, 2019

Taming emergent engineering

Understanding and engineering emergent behaviors is one of the long-standing challenges of complex systems.  Over the past fifteen years, one step at a time my collaborators and I have been pinning down the engineering of emergent behaviors.  Our most recent publication, however, represents quite a major step in the project.

"A Higher-Order Calculus of Computational Fields", out this week in ACM Transactions on Computational Logic, finally puts a solid mathematical link between collective phenomena and local actions.  In this paper, we present not one but two equivalent semantics for aggregate programs: one in terms of local actions of devices and the other in terms of collectives extending across space and time.  Every field calculus program expressed in one view can be automatically translated to the other, from global to local and from local to global.  We've been working with this result informally for many years, but now we have rock-solid mathematical proof.

Now combine that with "Space-Time Universality of Field Calculus", a paper we published last year demonstrating that every computable function over space and time can be implemented using field calculus.  That tells us that, no matter what emergent behavior you might be dealing with, if it is physically possible, there is guaranteed to be a program that can be expressed in our simple language that can both describe the collective behavior and be applied to produce it from local interactions.

This doesn't mean we can predict the behavior of any old system out there.  Just because you know there is a description doesn't mean it will be easy to find it, or that said description will be simple. Likewise, it might be difficult to understand the implications of a program. But having a simple language that is guaranteed to cover all of the relationships of interest can make a very big difference in just how hard that search space is to navigate.

Unfortunately, I don't really recommend that you read either paper unless you love wading through heavy mathematical symbology.  Ultimately, once you wrap your head around the mathematics, the core ideas of each paper are fairly simple and elegant, but there's a lot of supporting details that have to be dealt with, systematized, and pinned down with mathematical variable names.

Next step: making a more digestible summary of the key results available to the wider community who may be interested.

Example of resolving an aggregate function call over space and time in higher-order field calculus.