The Polish writer Stanislaw Lem (1921-2006) is one of the most influential science-fiction visionaries of all time. Mostly known for his novel Solaris (1961), which was later made into a film by Andrei Tarkovsky, Lem has been prolific in his fiction, often blending social satire with engineering fantasy. Space travel, human contact with alien intelligence and societies of robots are among his most favorite subjects. Lem has written non-fiction as well, such as the monumentalSumma Technologiae (1964), a reference to Summa Theologiae by Thomas Aquinas.
(Left: Stanislaw Lem (1921-2006))
In Summa Technologiae Lem speculates and philosophizes on prospective social, cybernetic and biological advances. What is of particular interest to AI is a chapter entitled “Intellectronics” where Lem discusses the concept of what he calls the “intelligence amplifier”.
The intelligence amplifier would be similar to existing machines that amplify human physical strength, for example cares, excavators, airplanes or cranes. A human connected via a suitable control system to an intelligent amplifier would be able to increase her IQ by many factors. Lem envisions intelligence amplifiers that could turn a person of average intelligence (100-110 on current IQ tests) to a super genius with an IQ of 10,000!
One could argue that intelligence amplifiers already exist – they are called “computers”. Computers increase our capacity for calculations manifold, with monumental and historical consequences in the way our civilization, our economy and our society evolves. They truly “amplify” our intelligence, collective and individual.
But Lem means something deeper than that. In comparing the IQ of the personwith the IQ of the person-plus-machine he suggests that the machine itself must be intelligent. In fact, just like a crane is more powerful than its human operator an intelligence amplifier must be more intelligent too!
The problem that Lem aptly identifies in the design of such a superintellgent machine is the obvious: the machine would have to be more intelligent than its designer. From a classical engineering perspective this means that one cannot develop design specifications (how can you describe what intelligence higher than yours can do and how it thinks?), and therefore one cannot even begin to imagine how this machine would be like, let alone how to control it. This is of course a paradox that strong AI faces when dreaming of “machine superintelligence”, i.e. machines smarter than humans that would, supposedly, usher us into the era of “AI Singularity”. One only wishes they will be friendly, or else…
Interestingly, in Intellectronics Lem suggests an exit from the superintelligence engineering conundrum. Instead of electronics he proposes the search for new substances, building materials which in certain aspects are similar to living organisms.
Forty-seven years after the writing of Summa Technologiae researchers in Caltech have invented a method for designing systems of DNA molecules whose interactions simulate the behaviour of simple artificial neural networks.
The researchers based their biochemical neural network on a simplified model of a neuron. The model neuron receives input signals, multiplies each by a positive or negative weight, and only if the weighted sum of inputs surpass a certain threshold does the neuron fire, producing an output. The model was built by synthesizing DNA strands in a test tube, i.e. a real “computing soup”!
Lulu Qian, the lead author of the paper (see reference) posed the main theoretical question of the research thus: “Instead of having a physically-connected network of neural cells, can a soup of interacting molecules exhibit brain-like behaviour?”
Although the invention is far from having practical applications any time soon, it could have a future in nanodrug design. More interestingly however, the Caltech invention explores notions of computability in living systems. As Qian explained in Science Daily, “Before the brain evolved, single-celled organisms were also capable of processing information, making decisions, and acting in response to their environment. The source of such complex behaviors must have been a network of molecules floating around in the cell. Perhaps the highly evolved brain and the limited form of intelligence seen in single cells share a similar computational model that’s just programmed in different substrates.”
It is a theory worth testing, and certain to have made Lem smile were he alive today. Unlocking the secrets of biological computation may indeed be the way to building the intelligence amplifier that the Polish novelist dreamt of, so many years ago.
Journal Reference: Lulu Qian, Erik Winfree, Jehoshua Bruck. Neural network computation with DNA strand displacement cascades. Nature, 2011; 475 (7356): 368 DOI: 10.1038/nature10262