It's nice to be noticed...

Well, this is interesting... New Scientist, Ars Technica, and various other places seem to have noticed one of our patent filings.  I guess that the combination of robots, 3D printing, and digital object capture hits the right combination of cool buzzword nerves (though somehow they all overlooked BBN Technologies in the rush to make Roomba jokes about our co-inventors at iRobot).

Indeed, it is pretty cool what can be possible when you bring the improving reasonable-cost capabilities in those areas together with the right sort of representations and AI techniques.  In fact, faithful readers, let me give you a better image than any of the filings had in them (damn patent drawing rules!).  Click to see it's full glory:

And that's enough said about that for the moment...

Operational Semantics of Proto

OK, OK, already, I'll post the Proto semantics paper.  Actually, technically, it's already been posted on my web page for a little while now, but now I'll actually point people at it and write about what is contained therein.

Honestly, I'm a little bit surprised that this has been requested.  But who am I to deny my colleagues the pleasure of what is, perhaps, the mathematically densest paper that I have ever written?  For those who may not be aware, the formal semantics of a programming language is the mathematical statements about what exactly all of those pieces of code mean.  And since programming languages are generally awfully complex, the formal statements of their nature are often even more so.

Except... if you do it really well and elegantly, like this Featherweight Java paper, or the R5RS manual for Scheme, then you can seize the vital core of the language in only a few operations and end up with only a page or two of dense mathematics.

This paper, though, we wanted to ensure we covered everything in the set of unusual operations that are made available and used by Proto, in its discrete approximation.  This is part of a general project on formalization of concepts in space-time computing that we've been conducting over the past couple of years.  We've produced formal continuous models, a prior key-operators-only discrete model, and now this full model of the discrete behavior of Proto.

It must be admitted, though, that the process of getting there has led to a paper that is sometimes colloquially referred to within my office as "the paper with the four different types of arrow."

I'm very happy with this paper, for all that I find our own work challenging to read without at least an hour to retrain the symbology into my brain.  I'm also quite grateful to my coauthor Mirko Viroli, for helping instigate this and leading on the semantics development.  We've ended up getting a much clearer focus on what's important and different about computations distributed over space-time: in particular, it turns out to be incredibly powerful to be able to constrain which regions of space a program runs in---and to implicitly pass that constraint down to all sub-programs within it.  We've also found a number of bugs and subtleties in our prior thinking, especially to do with boundaries in space and time.  And already, we are moving on, reducing these semantics to a simpler and more elegant form, that is not tied to Proto and where perhaps we can formally unify the continuous and discrete worlds... it is a hard project, but, I believe, an important one.

For those who swim well in these challenging seas... your thoughts are eagerly anticipated.

The Final Stop on My Northwestern Tour

Today, I helped Jimmy teach his day-long class on the smart grid to the group of ten students in his Energy course at BGI.  It's an interesting system, and I have no idea whether any other business schools are arranged this way or not.  Each class meets once a week, but then once a month they have an "intensive," where the entire student body goes together to this retreat in the woods and spends an entire weekend on day-long editions of their courses, as well as eating and socializing together.  It seems a wonderful bonding experience, and no doubt gives a lot of coherence to these cohorts of students.  Tonight, the students (and many of the faculty) will party together into the late hours of the night.

In between sessions, I explored the trails of the campus.  They snake through hundreds of acres of Northwest rain forest, around streams and bogs and ponds, with little classrooms embedded within them: a floating classroom that can be paddled across a lake, an elevated classroom raised up with the trees.  I enjoyed teaching, and I enjoyed my quiet time, drawing to the end of an intensive trip, just taking some bits of my weekend that will otherwise fail to be.

From Bellevue to Bainbridge

Today, I brought my message before a sort of audience that I personally have never faced before: the people who operate Snohomish PUD, one of the biggest publicly owned utilities in the country.  These folks are actually dealing with particular problems in particular substations in their territories, where technologies like the ones I work with could make a difference.  Different sorts of questions, different aspects I emphasize, but again, I came away very happy with our discussions and hopeful that we can work together.

The other thing that made me very happy about my day is that Seattle finally gave me a really proper fog.  For some reason, whenever I've been here in the past, it's been pretty much bright and beautiful and sunny---so unlike the way it's "supposed" to be.  Both last night and this morning, though, the fog was thick on the evergreens and wrapped everything in a chill damp softness, then froze to rime it with shining frost.

Now, I am in quiet dorm room with log-construction furniture on the other side of Puget Sound, after a ferry-ride away from a Seattle glowing darkly beneath low clouds that hid the tops of the buildings and turned the Space Needle into a truncated pair of arcs.  BGI holds its intensive weekend courses at an educational nature preserve called IslandWood.  In the dark, it smells of damp green and redwoods.  I took my shoes off as I entered the building, and walked up to my room past the great tall fireplace.  Now it is time to review my lecture notes and turn in for an early rest.

Stop #2: PNNL

Today was spent at Pacific Northwest National Labs, where I got to meet folks whose names I've known for a while, who did some of the best known prior work on demand response.  Once again, a damned good visit, with a lot of intense technical back and forth and possibilities for the future.

Beyond the professional part of the day, an interesting side note for me was seeing the Tri-Cities area again.  Richland, the location of PNNL, is where the Hanford Nuclear Reservation was set up during World War II, because they needed an isolated location to make plutonium and other radioactive nasties.  They're still there, slowly drifting towards the Columbia River, and in the years since, Richland has sprouted a great many organizations related to government and energy one way or another.  I was last there about 15 years ago, as an undergraduate on the MIT Aerial Robotics Team.  We were participating in a DOE-sponsored autonomous flying vehicle competition that was held next door to Hanford, at the training site for the firefighters who might have to deal with those horrifying hazards, and man did that place have the best props for a search and rescue contest, including 50-foot flame jets and tumbled train cars.

It was a huge amount of fun, going out there and working through the summer nights to try to get our helicopter finished and working properly.  In the end, we, like most of the teams, were never able to fly autonomously at all.  A number of teams couldn't even get off the ground---typically, the ones working on novel vehicle designs, rather than adapting a hobbyist vehicle.  The first year that I was there, we might have been able to, but we were felled by the heat: it was 100 degrees out, and much more right near the asphalt, and our helicopter had nylon gears driving its tail rotor, which softened and gave way while it was 30 feet in the air doing its startup checklist under remote control by our safety pilot.  The reason helicopters have tail rotors is to make sure the rotor spins and not the body: when the gears gave out, it came whirling down out of the sky, and it was only the great skill of our pilot that allowed him to autorotate it into a controlled crash.  The second year, we'd lost all of our good controls people to real, paying research projects, and the helicopter never got off the ground.  Fond memories and good lessons, but ultimately no success.

And yet: somehow, a paper came out of our failures and was published.  So far, it has gathered 4 citations.  Let that keep me humble as an author.

Pullman, and across the plains

My visit to Washington State University went extremely well.  I gave a good seminar to a good audience, and had a lot of interesting discussions with colleagues.  Lot of good opportunities for collaboration there, and that's the last I'll say on that for now.  And lucky you, dear reader, you have the opportunity to download my talk slides if you wish (pdf or pptx).  I'm quite happy with the presentation: I think it's a good framing of both the underlying problem, technical core, and the business context, and summarizes the current state of all my work on energy demand shaping.

Then I set off westward into the afternoon, to cover the 150 miles to Richland and meet my colleague Jimmy.  The sun angled sharply down across the fields, where a thin frosting of snow turned the hills to corduroy and whorls.  As I climbed out of the valleys, away from the Palouse river, the land emptied and the other cars vanished, clipping along up to the vast plains.  On the last dregs of cell phone signal, I heard about my daughter's first attempts at crawling and bade my wife goodnight, then pushed on into a lowering mist, bluegrass carrying me across the emptiness into the gathering dusk.

Perhaps this isn't why I became a scientist, but it's a wonderful bonus from the job.

Beginning my Northwest Energy Tour

Back in November, when I spoke on a smart-grid panel in Seattle, my trip was a very productive introduction to other folks working in the area here in the Northwest---thanks in large part to my old friend and colleague Jimmy.  It's funny how the two of our careers have weaved in and out of contact in the area over the last few years, but I'll leave that story for another time.

In any case, I'm back in Washington state again, for a series of visits and talks, the schedule anchored around a guest lecture in Jimmy's business class at BGI (a very funky sustainable business school / startup incubator in Seattle, that I found pretty awesome when I visited it last time).  The (somewhat intense) itinerary is:

• Washington State University, in Pullman - home of a lot of world class power-engineering research
• Pacific Northwest National Labs, in Richland - one of the epicenters of the smart grid
• Snohomish PUD, in the Seattle Area - one of the most innovative public utilities
• Bainbridge Graduate Institute, in Puget Sound - helping teach Sustainable Energy Solutions

So, I will be going pretty much end-to-end in Washington state over the next four days, talking nothing but energy with all I meet, and at the end, shoveling myself on a plane back home.  It will be intense, and hopefully quite productive in future collaborations.

And now, having just returned from a nice dinner with my colleagues at WSU, and pushing well past midnight EDT, it is time to head to bed.

The Measurement of Babies

Sorry, for anyone who may be bored of it, but I've got to post another science/baby post.  I was planning to (finally) write about our programming language semantics paper that came out a few weeks ago, but I'm deferring it again because I'm just more psyched to write about baby-metrics right now.  So you'll just have to wait another week before you hear about the most symbol-dense paper I have yet published.

As anyone who is a parent surely knows, tracking the progress of one's child can become a minor obsession.  There are just so many resources available that will happily tell you what to expect at each month, what's normal, what's not, etc.  I've fought back against this by embracing the Dr. Spock approach, which basically boils down to: "First, enjoy your baby.  Second, learn from your baby."

Still, there are a few things that you can't help being informed of at regular intervals, like height and weight percentiles.  Harriet had a rough start, and in her first few days dipped down near the 5th percentile weight line (the level at which Medicine Becomes Concerned), so we are perhaps even more aware of those than we might be otherwise.  She caught up well, so no worries.  But now in the long interval between her 4-month and 6-month checkups (hah!), we started wondering whether our seemingly always ravenous little girl was going overboard on the other side, and ended up looking up height and weight charts.

No problems, and I basically forgot about it until today, when Harriet was playing with a carpentry level (another fine household object toy), and we realized she was taller than it---more than 30 inches, which according to my chart was well above 95th percentile for her age! Well, nothing gets a couple of scientists going like an unexpected result... out came the hypotheses.

Could the chart be wrong? Careful inspection found a CDC watermark and url blurred in the bottom of the image, which led to the original rather trustworthy source.  There is even, delightfully, a complete methods paper, which I pulled up and started going through, while the subject of the investigation shifted her attention to a joint manual/oral assay of the texture of socks (Fabric Sciences 102).

Yes, our baby is apparently a long, trim little girl, according to primary sources.  But what was their statistical data pool?  Do we know if there are ethnic variations?  What would that say about an F1 hybrid of Anglo-Saxon and Bengali?  Are there epigenetic effects linked to immigration to a high-nutrition country?

The CDC report itself didn't have enough data for studying the extremas of different ethnicities: although their birth statistics were massive, the infant-to-toddler range only comes from a few tens of thousands of data points, and the data is all at least 15 years old, when the country was significantly more homogeneous.  Their discussion and citations on the matter, though, argue very solidly that ethnicity has pretty much null effect compared to nutrition---it looks like there's been a lot of research on the matter, and especially regarding Mexican-Americans.  I haven't been able to find any references specifically investigating biracial children yet... but we didn't dig very deeply, since a real baby has priority over a metrized baby-model.

And the conclusion is... we have a long, trim little girl, and her parents can just leave the rest be, unless we want to have fun telling Just-So stories.

Materialism and the Six Month Old Child

In the wake of Harriet's first Christmas, my wife and I were organizing her stuff and realized just how much Baby Material Goods have steadily flowed into our lives over the past six months.  This is not just the necessaries, but lots of extra clothes and toys and stuffed animals, enough that she has barely played with many of them.  Most of this is not things that we ourselves bought, but gifts from friends and relatives---each individually quite appreciated and carefully thought out, but aggregating together into an overload of material goods.

And yet, with all these lovely, carefully-crafted-for-babies, well-thought gifts, do you know what her current three absolute favorite toys are?

1. The caps of her baby food containers (great for dexterous manipulation practice!)
2. A plastic clothes hanger (fascinating and challenging geometry)
3. Our guava tree (strictly limited by her parents, to prevent damage to the tree)

I think this says something very interesting about infant cognition, which shouldn't be a surprise to anybody who has followed the literature or has read popularizations like The Scientist in The Crib.  I didn't see it myself until my daughter rubbed my nose in it, though.  It seems that many modern baby toys are surprisingly uniform in their color (highly saturated) and material properties (from plush/felt to firm plastic, with optional crinkling).  Harriet, however, is often much more interested in ordinary household objects, and we have generally felt that the best way to encourage her active, curious, and playful nature is to help her to get at the things that catch her interest, so far as they are safe.

I think that part of this is her being interested in the things that adults pay more attention to, part of it is because of the "forbidden fruit" aspect of some of them (especially our guava tree).  More than anything else, though, I would guess that our little learning engine just finds the physical properties of these household goods much more interesting.  When you stop and really look at the things around your home, that is unsurprising: many of them are really interesting composite materials.  For example, she really likes blankets, which have this non-linear behavior where they bunch when you push on them, but stretch slightly and then become stiff when you pull on them, not to mention the propagating folds and ripples from the fact that they are a two-dimensional sheet.  Another recent investigation was of a net-covered box of clementine oranges, where you can put your hands under the net and see them through the holes even as the (decidedly non-trivial) manifold constrains your ability to move them.  And her plastic clothes hanger is a decidedly complex geometric shape, which can catch on things and constrain motion in lots of different ways---say, by looping around your arm when you're pushing on the flat base, so that when you push it, it pushes you from behind.

Watching her investigating these things reminds me how much we take our intuitions about these materials for granted, and yet they are often fantastically complex from a physics and materials science perspective.  If I remember correctly, I believe that the physics of fabrics and of foods (so many complex colloids!) still holds quite a number of unsolved scientific problems.  Should I be surprised that our infant finds such complex and tactilely stimulating objects fascinating?

So now we are simplifying with respect to baby toys, and letting our little darling spend more time investigating the ordinary things in her world, which are really not so ordinary.  When she finally figures out crawling, it will be both a blessing and a curse: our next physics investigation is a lowering of the center of gravity of our bookcases, in anticipation of that day...