Better than the Turing test

Is the Turing test really the best way to screen for artificial intelligence? One alternative sounds more promising: Winograd schemas.

The test would take the form of a multiple-choice quiz of reading comprehension. But the text itself would have some very specific features. It would consist of Winograd schemas: pairs of sentences whose intended meaning can be flipped by changing just one word. They generally involve unclear pronouns or possessives. A famous example comes from Stanford computer scientist Terry Winograd:

"The city councilmen refused the demonstrators a permit because they feared violence. Who feared violence?"
1) The city councilmen
2) The demonstrators

And:

"The city councilmen refused the demonstrators a permit because they advocated violence. Who advocated violence?"
1) The city councilmen
2) The demonstrators

Most human beings can easily answer these questions. We use our common sense to figure out what "they" is supposed to be referring to in each case. And that common sense basically involves a combination of extensive cultural background knowledge with analytical skills. (In the first question, we can deduce that the city councilmen feared violence. In the second, the demonstrators advocated violence.)

For computers, however, these questions can be quite difficult. From a grammatical standpoint, the "they" in the sentences is technically unclear. In both questions, "they" could be either the councilmen or the demonstrators.

Is this truly a superior test of AI? We may start collecting some data points soon. In 2015, Nuance Communications is sponsoring the first of what will be an annual Winograd Schema Challenge.

I continue to insist that the next great milestone in AI is when Waze stops trying to send me to unprotected left turns as shortcuts. UPS already figured this out.

First digital animal

One approach to building artificial intelligence is to just build a perfect copy of an animal intelligence (including humans). We are a long ways away from being able to do that. How long I have no idea, but safe to say it's not imminent.

But someday we may look back on this digital worm as a significant milestone in the attempt to replicate a live organism without using genetic cloning.

The bot's artificial brain has the same number of cells as a real nematode brain, and they are connected up in exactly the same way. But instead of a fluid tubular body animated by 95 muscles, WormBot has a plastic body and two wheels. It does not eat, defecate, reproduce or die. That will be left to its future sibling, WormSim, which will be a cell-for-cell digital copy of the worm, living inside a computer.

Both projects began with the simplest, smallest brain that we know of – the one that is inside the nematode worm Caenorhabditis elegans. This lab workhorse was the first organism to have its genome sequenced, and the first to have its entire brain mapped. It is largely hermaphrodite, with 959 cells each of which has also been mapped. Its network of 302 neurons connect via 6393 synapses – its connectome – and link to the worm's 95 muscles at 1410 junctions.

...

Independent researcher Tim Busbice used the OpenWorm data to build WormBot's brain. He started by building a neural network in which the neurons were connected to each other according to the C. elegans connectome. In a live animal, neurons often latch onto each other through repeated connections: so neuron A might make five synapses to neuron B, and each time A fires, all five relay the signal to B. Busbice used weightings to represent this in his neural network. Just like real neurons, those in WormBot, receive inputs from a network of "upstream" neurons and have to reach a threshold in order to fire and pass the signal on to those downstream. If that threshold isn't reached a set period of time the system resets to zero.

Instead of muscles, WormBot has two wheels controlled by a matrix of 95 cells, representing the 95 muscles of C. elegans. Busbice hooked up the worm's chemosensory neurons, which a real worm uses to detect smells and tastes, to a microphone that is triggered beyond a certain decibel threshold. He also connected the worm's "nose-touch" neurons to a sonar that would send a message upstream to the brain if WormBot gets within 20 centimetres of an obstacle.

Assuming you can create a perfect copy, profound philosophical questions of consciousness, the soul, and all that are sure to follow.

Becons

The Becons are the Behavioral Economics Oscars. They are exactly what they sound like, prizes awarded on the basis of their economic lessons. Cass Sunstein runs down this years winners.

Best picture: No, it’s not "Interstellar," and it’s not "Gone Girl." And a loud system 1 rejection of "Birdman," "The Theory of Everything," "Into the Woods" and "The Imitation Game." The biggest Becon goes to the movie that has the biggest heart, and the best scene, and the best score, and the best romance of the year (without even a single kiss). It’s a celebration of optimism bias, the value of agency, duration neglect, the illusion of control, the gambler’s fallacy, steps you can’t take back, and the human spirit. Mark Ruffalo and Keira Knightley, you are ridiculously good. "Begin Again" dances off with the Becon.

Which goes to show that while movies can teach important economic lessons, that might not be the best way to go about making a good movie (Begin Again is by the director of Once, and it is basically Once with actors instead of musicians as leads; if you choose to see one, see Once).

Aliens is the best movie about humans and technology

I love this Tim Carmody essay about why James Cameron's Aliens is the best movie about technology.

Now, the aliens are certainly intelligent enough to use tools — they just don’t need them. Evolution has given them everything they need to kill just about anything they want to. They can gestate inside any host, taking on whatever physical characteristics of that host are needed to survive in a new environment. Their bodies make their armor, their secretions make their architecture. Technology is irrelevant. Their biology is so perfect that they are technology. (I’m partial to the theory that the derelict spacecraft discovered in Alien and again in Aliens was carrying the alien’s eggs to use as bombs; Burke and the Weyland-Yutani corporation want to exploit their biology in much the same way. It’s all IP to humans!)

...

The climactic scene in Aliens is obviously Ripley’s battle with the alien queen. Exoskeleton to carapace, blowtorch to hidden extra jaw, it’s literally a battle between a human plus technology and a biologically superior lifeform. Again, I love how low-tech Ripley’s suit actually is: it could easily have existed in 1986, but probably never will, if Amazon’s warehouse robots are any indication of our future. But I also love that Ripley wins not just because she can put on the suit and know how to use it, but because she can take it off.

Think about it: all Ripley is really doing inside the exoskeleton is holding off the queen while she is opening the airlock. Even with the benefit of a robotic prosthesis, a human being can’t kill the alien: only space can. Ripley and the queen tumble head over head into the bottom of the airlock. Ripley luckily lands on top, but is also able to ditch her techno-suit and climb out of harm’s way. The alien queen can’t even break off from her egg sac as quickly as Ripley scrambles out of her suit; she’s too tied to her own biology to let any of it go, to treat it as something other than herself.