2006 : WHAT IS YOUR DANGEROUS IDEA?

stanislas_dehaene's picture
Neuroscientist; Collège de France, Paris; Author, How We Learn
Touching and pushing the limits of the human brain

From Copernicus to Darwin to Freud, science has a special way of deflating human hubris by proposing what is frequently perceived, at the time, as dangerous or pernicious ideas. Today, cognitive neuroscience presents us with a new challenging idea, whose accommodation will require substantial personal and societal effort — the discovery of the intrinsic limits of the human brain.

Calculation was one of the first domains where we lost our special status — right from their inception, computers were faster than the human brain, and they are now billions of times ahead of us in their speed and breadth of number crunching. Psychological research shows that our mental "central executive" is amazingly limited — we can process only one thought at a time, at a meager rate of five or ten per second at most. This is rather surprising. Isn't the human brain supposed to be the most massively parallel machine on earth? Yes, but its architecture is such that the collective outcome of this parallel organization, our mind, is a very slow serial processor. What we can become aware of is intrinsically limited. Whenever we delve deeply into the processing of one object, we become literally blind to other items that would require our attention (the "attentional blink" paradigm). We also suffer from an "illusion of seeing": we think that we take in a whole visual scene and see it all at once, but research shows that major chunks of the image can be changed surreptitiously without our noticing.

True, relative to other animal species, we do have a special combinatorial power, which lies at the heart of the remarkable cultural inventions of mathematics, language, or writing. Yet this combinatorial faculty only works on the raw materials provided by a small number of core systems for number, space, time, emotion, conspecifics, and a few other basic domains. The list is not very long — and within each domain, we are now discovering lots of little ill-adapted quirks, evidence of stupid design as expected from a brain arising from an imperfect evolutionary process (for instance, our number system only gives us a sense of approximate quantity — good enough for foraging, but not for exact mathematics). I therefore do not share Marc Hauser's optimism that our mind has a "universal" or "limitless" expressive power. The limits are easy to touch in mathematics, in topology for instance, where we struggle with the simplest objects (is a curve a knot… or not?).

As we discover the limits of the human brain, we also find new ways to design machines that go beyond those limits. Thus, we have to get ready for a society where, more and more, the human mind will be replaced by better computers and robots — and where the human operator will be increasingly considered a nuisance rather than an asset. This is already the case in aeronautics, where flight stability is ensured by fast cybernetics and where landing and take off will soon be assured by computer, apparently with much improved safety.

There are still a few domains where the human brain maintains an apparent superiority. Visual recognition used to be one — but already, superb face recognition software is appearing, capable of storing and recognizing thousands of faces with close to human performance. Robotics is another. No robot to date is capable of navigating smoothly through a complicated 3-D world. Yet a third area of human superiority is high-level semantics and creativity: the human ability to make sense of a story, to pull out the relevant knowledge from a vast store of potentially useful facts, remains unequalled.

Suppose that, for the next 50 years, those are the main areas in which engineers will remain unable to match the performance of the human brain. Are we ready for a world in which the human contributions are binary, either at the highest level (thinkers, engineers, artists…) or at the lowest level, where human workforce remains cheaper than mechanization? To some extent, I would argue that this great divide is already here, especially between North and South, but also within our developed countries, between upper and lower casts.

What are the solutions? I envisage two of them. The first is education. The human brain to some extent is changeable. Thanks to education, we can improve considerably upon the stock of mental tools provided to us by evolution. In fact, relative to the large changes that schooling can provide, whatever neurobiological differences distinguish the sexes or the races are minuscule (and thus largely irrelevant — contra Steve Pinker). The crowning achievements of Sir Isaac Newton are now accessible to any student in physics and algebra — whatever his or her skin color.

Of course, our learning ability isn't without bounds. It is itself tightly limited by our genes, which merely allow a fringe of variability in the laying down of our neuronal networks. We never fully gain entirely new abilities — but merely transform our existing brain networks, a partial and constrained process that I have called "cultural recycling" or "recyclage".

As we gain knowledge of brain plasticity, a major application of cognitive neuroscience research should be the improvement of life-long education, with the goal of optimizing this transformation of our brains. Consider reading. We now understand much better how this cultural capacity is laid down. A posterior brain network, initially evolved to recognize objects and faces, gets partially recycled for the shapes of letters and words, and learns to connect these shapes to other temporal areas for sounds and words. Cultural evolution has modified the shapes of letters so that they are easily learnable by this brain network. But, the system remains amazingly imperfect. Reading still has to go through the lopsided design of the retina, where the blood vessels are put in front of the photoreceptors, and where only a small region of the fovea has enough resolution to recognize small print. Furthermore, both the design of writing systems and the way in which they are taught are perfectible. In the end, after years of training, we can only read at an appalling speed of perhaps 10 words per second, a baud rate surpassed by any present-day modem.

Nevertheless, this cultural invention has radically changed our cognitive abilities, doubling our verbal working memory for instance. Who knows what other cultural inventions might lie ahead of us, and might allow us to further push the limits of our brain biology?

A second, more futuristic solution may lie in technology. Brain-computer interfaces are already around the corner. They are currently being developed for therapeutic purposes. Soon, cortical implants will allow paralyzed patients to move equipment by direct cerebral command. Will such devices later be applied to the normal human brain, in the hopes of extending our memory span or the speed of our access to information? And will we be able to forge a society in which such tools do not lead to further divisions between, on the one hand, high-tech brains powered by the best education and neuro-gear, and on the other hand, low-tech man power just good enough for cheap jobs?