SMARTSUpdate: We, the Disposables, and DARPA

Okay, I confess, I was bored the other day. What to do?

Peruse DARPA's website, I said to myself, that's always good for a shock to the head. 

Someone revised it since the last time I looked. It's much more approachable now (yet no more open). There are at least 200 projects running or planned. I scroll and scroll and scroll. One that catches my eye is about teaching computers to understand and communicate using human languages. NSA and all the other agencies in Five Eyes would be thrilled to have something like that, an artificial intelligence able to understand all the cell conversations they've been hauling in. Think of the savings: they could get rid of all those annoying young analysts who take up space and insist on getting paid and sometimes walk out of contractors`offices with the Crown Jewels.

Another project concerns a system for remote sensing across the Arctic, a system that would vastly reduce the need for human surveillance.

Wait a minute, I say to my screen, most of the Arctic belongs to Canada, you Americans only have Alaska.  Canadian Inuit in the ranger program stand on guard for us, so what the hell is this?  I am relieved to see that one member of the DARPA team involved in this project is a man who grew up just down the street from me, a Canadian (though he is now a learned professor at an American university). With Jeffrey on the case, I surely don't have to notify my MP, the new Canadian Minister of Indigenous Affairs, about DARPA's plan for smart machines on our turf, in our Arctic waters? Or do I?

I keep scrolling. Another project pops up that stops my breath. It's called the ICARUS program.

DARPA, like other military institutions, displays such a grim determination to spew forth acronyms no one can figure out that serve some obscure PR purpose. ICARUS, I read, stands for " Inbound Controlled Air-Releaseable Unrecoverable Systems."  The acronym is meant to bring to mind the story of Icarus, the Greek boy who flew on wings made from feathers and wax which melted when he got too close to the sun. Icarus's adventure ends badly in the ocean.

See what I mean about obscure PR? This seems like a very odd choice for a project name: normally, one would want one that suggests an upside in return for the expenditures of public funds. Something tells me that DARPA's version of the Icarus story is a tad skewed, possibly because they stripped it down for use in their narrative software project (see previous blogpost). Here's what I found in the online encyclopaedia of Greek Mythology.

" Son of Daedalus who dared to fly too near the sun on wings of feathers and wax. Daedalus had been imprisoned by King Minos of Crete within the walls of his own invention, the Labyrinth. But the great craftsman's genius would not suffer captivity. He made two pairs of wings by adhering feathers to a wooden frame with wax. Giving one pair to his son, he cautioned him that flying too near the sun would cause the wax to melt. But Icarus became ecstatic with the ability to fly and forgot his father's warning. The feathers came loose and Icarus plunged to his death in the sea."

Icarus. (zoom)

Why would DARPA invoke doomed Icarus instead of clever Daedalus? Because, the point of the Icarus story is that the wings disappear. What DARPA wants to build is a small, low cost, disposable "aircraft".

"The objective of the ICARUS program," DARPA explains, "is to create a vanishing platform for the airborne delivery of small payloads. Supply and re-supply of small military and civilian teams in difficult-to-access territory currently requires the use of large, parachute-based delivery systems that must be packed out after receipt of the payload both for operational security and environmental concerns...."

Last month, DARPA proposers were invited to a meeting on ICARUS, a meeting closed to the media. (The media of course visit the DARPA website and so a number of stories about disappearing drones soon made their appearance in the usual locations.) The parameters of the project were laid out as follows: DARPA asked interested parties to design, prototype and demonstrate "autonomous, air delivery vehicles capable of gentle...delivery [of] a three lbs. payload with 10 m accuracy to a GPS-programmed location. Upon payload delivery the vehicle must rapidly physically vanish, i.e., the vehicle's rapid physical disappearance, or transience, is part of its mission specific."

Well wow, I said to myself. Talk about planned obsolescence. What sort of material exists out there that can be fashioned into a plane that will carry a load while making its way without human aid to a particular location, and then just disappear? Doesn`t that sound more like a genie than a technology? And yet, DARPA explained, the ephemeral part of the problem has already been solved in a previous DARPA program with an equally silly acronym, VAPR, which stands for Vanishing Programmable Resources Program. VAPR " developed self-destructing electronic components" and other kinds of vanishing materials such as "structurally sound transient materials....[including] ephemeral... polymer panels that sublimate directly from a solid phase to a gas phase, and electronics-bearing glass strips with high-stress inner anatomies."

These glass strips, when hit with the right radio frequency, shatter into a lot of tiny wee particles of silicon. IBM has been issued a DARPA contract to make the electronics under glass.

Why, you ask, does DARPA want these things?

" A goal of the VAPR program is electronics made of materials that can be made to vanish if they get left behind after battle, to prevent their retrieval by adversaries."

Of course the products designed in the ICARUS and VAPR programs will have many other uses beyond military applications.  Amazon would love to have a cheap little drone that could make its way autonomously to your doorstep, deposit your book, and then disappear instead of having to return to base.  Detroit would be thrilled if it could make cars disappear on command or merely stop working by means of a radio broadcast: that would really drive sales, wouldn't it? The security agencies would appreciate autonomous flying machines that can deliver something unpleasant to an unwitting enemy and leave no trace of origin. And never mind the security agencies, consider the Mob. They already use drones to drop goodies like drugs and cellphones into prison compounds.  Disposable drones would be so much better. And by the way, if you can make a disposable autonomous drone from a polymer, why not a gun brought to you by a 3D printer, that can shoot, kill, and then disappear?

Which brings me to disposable humans. There is a certain underlying aim to these DARPA programs which reaches its apotheosis in one more project that caught my eye. They all involve eliminating the need for human effort, in fact, for humans. The Big Mechanism project takes this to a whole new level. It was begun last year, so if you heard about it then you've probably forgotten about it by now, which is part of the problem it might well solve.

Here's what it's about.

The idea is that science, that grand edifice of new knowledge, produces way too many discoveries for anyone to keep track of. This is why specialties evolved, yet nowadays,  specialist scientists cannot manage the information flow in their own fields let alone discoveries made in other areas that might be helpful. So DARPA thought: what if we make a machine intelligence that keeps track of all the data sets out there in a certain area, a machine smart enough to read everything in the literature and to note and to cross collate the important information in all published papers so as to discern patterns of meaning, which will lead to experiments, which will lead to important discoveries? DARPA's team leader in this area, Dr. Paul Cohen, a computer scientist with a special interest in intelligent systems, believes that the causes of disease will be found in this way-- the big mechanisms, not just the relationships or associations we derive now from big data.

The Big Mechanism project will pay specific attention to everything published with regard to a certain gene group that might play a role in-- ta dah!-- cancer.

The point here is to machine the practice of science, to push beyond what mere humans can achieve. If you think I'm exaggerating, consider this ad for post docs to help with this work posted by University of Chicago's Knowledge Labs (found on Google). The job is described as: "post docs will work with a cross-disciplinary team of computer scientists, linguists, biologists, applied mathematicians and social scientists to build an automated system that extracts and integrates information from the literature on cancer biology, electronic medical records, and experimental data; models the evidence and assigns confidence to each cancer-related claim; then builds algorithms to reason over that data and propose new hypotheses; which will subsequently be sent to automated ' robotic'  experiments."

Last week I described how the professions are under pressure from onrushing developments in artificial intelligence. Someone challenged me about that: what about original thought, one person asked in a comment on Google+.  He meant that in his view it was unlikely that a machine intelligence could ever be original, could ever replace that spark of human insight that has led us from scratching obscure marks on cave walls in southern Africa to all that we can do today.

You can't get more original, more insightful, than Darwin or Einstein.

This Big Mechanism system is aimed at doing what they did, better.

If it succeeds, humans in science will also become disposable, in favor of machines.

This will not be an unintended consequence.

Does anyone in the US government ask DARPA to show why such developments are in the public interest?

Most of us have no idea how far and how fast this is moving until we start looking.