Proof from Intelligence (3)


Problem Solving

In the meantime, let’s start with a big one: problem solving. We pride ourselves on being able to fix situations that wouldn’t occur naturally. The lives of the Apollo 13 astronauts depended on fitting a square carbon dioxide filter into a smaller round hole, otherwise they would suffocate on their own breath. Nothing in that scenario is natural.

Wire isn’t natural, either, yet a crow surprised us by bending it into a hook. New Caledonian crows have been making hooks for some time, actually, but in the wild they use twigs instead. They learn this trick by watching other crows do it, too, and not by figuring out for themselves. For one of them to bend a material they’d never seen before, in a way they’ve never witnessed, is a significantly harder problem.

Betty pulled it off on her first attempt.

Jackie Chappell and company didn’t mean to test that. Betty and Alex, a male crow, were presented with a collection of bent and straight wire, then tested to see if they could use them to grab a tasty treat that was otherwise out of reach. Wire was used because both crows had rarely seen it in their lives. When Alex nabbed the only bent wire and flew off, Betty grabbed a straight piece and bent it into a hook with her beak and feet.

After the researchers picked themselves up off the floor, they devised a new test. Separately, each crow was given a straight wire and a treat that was otherwise out of reach. Out of ten trials, Alex only succeeded once, and even then he cheated. Nine times out of ten, Betty tried to pick up the treat with the straight wire, failed, then bent the wire into a hook using her beak, feet, or the tube containing the food, and succeeded in getting the treat.

Betty had be raised in captivity, so she couldn’t have learned this trick from her wild peers. She had to analyse this new situation, find a solution using what she knew about the materials and herself, then put it in motion. That’s novel problem solving, done in a species with a much smaller brain than ours.

There’s also the case of a feisty octopus at the Sea Star Aquarium in Coburg, Germany. It did not like captivity one bit, and found creative ways to protest. Otto would juggle its non-mobile tank-mates, sometimes hiding them under grate covers, and several times shorted out a bright light by squirting it.

Think about that last one. Octopuses live entirely underwater, where they swim by sucking in and squirting out water. Light bends when it moves from air to water or vice versa, and even we humans take a fair bit of training to compensate for that. So in order to hit that light, Otto had to take an organ it uses for a single purpose and put it to a different use in a habitat it never visits with physics quite different from its home turf, and hit a small target that isn’t where it appears to be.

Nothing about that is natural, either.

Mathematics and Logical Thinking

Neither is calculus, for that matter. And yet human beings have no problems doing complicated arithmetic in their heads, or pondering long chains of subtle logic. Our fellow species can barely count, in comparison.

There is one crucial difference, however: Homo Sapiens Sapiens goes to school. We’re not born math wizards, we have to be taught via long, intensive training sessions. Remove those, and our huge advantage goes with it. Good proof of this comes from the languages of hunter-gatherers. They spent most of their time sleeping, doing chores, gathering food, or fighting. There was no time or need to invent mathematics, so whatever number systems they came up with reflect our uneducated understanding of number.

Their achievements are depressing. Many hunter-gatherers could barely count, usually reaching no higher than one or two before invoking words that mean “few” or “many.” Some didn’t even have the concept of “one”:

In Pirahã, there are two words which prototypically mean ’one’ and ’a couple’ respectively, but it has been checked fairly extensively that their meanings are fuzzy ’one’ and ’two’ rather than discrete quantities (Everett 2005, 2004, Frank et al. 2008). It is not possible to combine or repeat them to denote higher (inexact?) quantities either (Gordon 2004). The Pirahã have the same cognitive capabilities as other humans and they are able to perform tasks which require discerning exact numeration up to the subitizing limit, i.e. about 3 (Gordon 2004). They just do not have normed expressions even for low quantities, and live their life happily without paying much attention to exact numbers.


(Unsupervised Learning of Morphology and the Languages of the World,” chapter Nine. Harald Hammarström , 2009)

The last two sentences of that quote bring up more evidence; our subitizing limit, better known as our working memory capacity, is only three or four items.[44] If you’ve had no training on how to count or do math, that’s the only storage space you have for numbers, and thus it limits how high you can count.

Interestingly, our species’ subitizing limit is on par with other species.

In a study published last summer in the Proceedings of the Royal Society B, Kevin C. Burns of Victoria University of Wellington in New Zealand and his colleagues burrowed holes in fallen logs and stored varying numbers of mealworms (beetle larvae) in these holes in full view of wild New Zealand robins at the Karori Wildlife Sanctuary. Not only did the robins flock first to the holes with the most mealworms, but if Burns tricked them, removing some of the insects when they weren’t looking, the robins spent twice as long scouring the hole for the missing mealworms. “They probably have some innate ability to discern between small numbers” as three and four, Burns thinks, but they also “use their number sense on a daily basis, and so through trial and error, they can train themselves to identify numbers up to 12.”

More recently, in the April issue of the same Royal Society journal, Rosa Rugani of the University of Trento in Italy and her team demonstrated arithmetic in newly hatched chickens. The scientists reared the chicks with five identical objects, and the newborns imprinted on these objects, considering them their parents. But when the scientists subtracted two or three of the original objects and left the remainders behind screens, the chicks went looking for the larger number of objects, sensing that Mom was more like a three and not a two. Rugani also varied the size of the objects to rule out the possibility the chicks were identifying groups based simply on the fact that larger numbers of items take up more space than smaller numbers.

(“More Animals Seem to Have Some Ability to Count,” by Michael Tennesen. Scientific American, September 2009.)

We’ve managed to out-reason other species because we found a very efficient way to gather food, which freed up enough spare time to come up with wonderful systems of math, and because our longer lifespans increased the odds of us stumbling on a technique, or gave us more time to learn it from someone else. No other species has pulled off both feats; elephants and whales rarely use tools to gather food, and wild crows only live eight years.

When you provide both time and training, other species can break past the subitizing limit too.

[Pepperburg] discovered that Alex could accurately add two sets of objects, such as crackers or jelly beans, so long as the total was six or fewer. In related work, Alex learned to order the Arabic numerals 1 through 8 (in the form of multi-coloured refrigerator magnets) in the correct order. She says he then spontaneously learned to equate these symbols with the appropriate number of objects.

In the newly published work, Pepperberg tested whether Alex could correctly add the Arabic numerals and also whether he could sum three sets of objects totalling 6 or less. Both experiments were cut short when Alex died, but Pepperberg says that the parrot did better than chance in both experiments.

In 12 trials of the Arabic numeral addition task, when asked “How many total?” he indicated the correct sum 9 times, demonstrating that 3 + 4 is 7, 4 + 2 is 6, 4 + 4 is 8 and so on. When presented sequentially with three sets of objects hidden under three cups, and asked how many, Alex offered the correct answer eight out of 10 times. He determined, for instance, that one, two and one jelly beans adds up to four.

(“Alex the parrot’s last experiment shows his mathematical genius,” Ewen Callaway. Nature News Blog. )

Even if you don’t agree with the above argument, there’s still the mechanistic one. As I write this, the fastest computer in the world can perform about 8,162,000,000,000,000 math operations per second, to sixteen digits of precision. The computer I’m typing this document on can manage roughly 1,570,000,000, and even my phone does 6,900,000. In comparison, try working out this slightly easier calculation entirely in your head:

 

29669907

x

42669080

 

 Currently, Marc Jornet Sanz is the fastest multiplier on this planet. He can do the math above in about thirty seconds, without any mechanical aids, which translates to roughly 0.04 calculations per second.

Computers can do more than mundane arithmetic, too. Mathematicians have begun to rely on them for proving theorems. They are commonly used to verify proofs, a tedious and error-prone task, but computers are increasingly generating their own proofs. To name one example, the Robbins conjecture was proven by EQP, a computer program developed at Argonne National Laboratory in the United States.

If mathematics and logic can be done as well, or even better, by a machine, we have no reason to think of them as gifts from a god.


[44] Thanks to a misunderstanding, most people think this number is actually seven. See “Seven plus or minus two,” by Jeanne Farrington. Performance Improvement Quarterly, 23: 113–116.