The Technology Review posted the other day an article on the “10 Emerging Technologies of 2008“. Very promising technologies indeed, but the one that really caught my attention was the Probabilistic Chips currently studied by Krishna Palem.
The reason I find this particular research so interesting is, that
- it has the potential of extending current scaling laws - and particularly Moore’s Law
- it takes advantage of the principles of collective intelligence and the wisdom of crowds.
The theory of Probabilistic Chips
According to the article
Palem has developed a way for chips to use significantly less power in exchange for a small loss of precision. [...] chips could be designed to produce the correct answer sometimes, but only come close the rest of the time. Because the errors would be small, so would their effects: in essence, Palem believes that in computing, close enough is often good enough.
Current Scaling Laws
From the beginning of the twentieth century until today, a series of laws has emerged concerning the growth of computers and networks. They all negotiate the value of a network according to the technological leverage of the time.
- Sarnoff’s law, which was interested in the growth of radio and television networks (the value of the network is proportional to the number of actors)
- Metcalfe’s Law describing the value and the growth of small scale networks (the “value” or “power” of a network increases in proportion to the square of the number of nodes on the network).
- Reed’s Law describing the value of Group Forming Networks (the value of networks, that support the construction of communicating groups create value that scales exponentially with network size).
- Moore’s law handling the evolution and expansion of computer microchips (the number of transistors on a chip will double about every two years).
One can notice the evolution from Sarnoff’s Law to Reed’s Law.
While Sarnoff’s Law was suggesting, that the value of a broadcast station (television or radio) would increase proportionally to its audience it could not be applied to more complex networks, since the degree of interconnectivity was much higher.
So Metcalfe’s Law was an evolutionary step of Sarnoff’s Law, which better described the upcoming computer networks (ARPANET) in the 1960s. But Metcalfe’s Law could be easily be applied to small computer or telephone networks, but certainly not to huge networks like the Internet.
Reed’s Law emerged to describe the masive development of the web and more specifically of the social web.
Moore’s Law, on the other hand remains stable. Although it finds application on the expansion of computer microchip technology for more than 40 years, its validity is lately at stake. The reason is, that silicon transistors are becoming smaller and smaller and therefore less reliable.
But that’s where the probabilistic chips come in play, to keep the microprocessor technology rapidly evolving.
With probabilistic chips, tiny microprocessors may be designed in such way, that the individual parts might be imperfect, but collectively bring perfect results.
Collective Intelligence
The probabilistic chip technology actually takes advantage of the theory of collective intelligence.
As supported by the Condorcet Jury Theorem [pdf], the probability of a correct answer by a majority of the group increases toward 100% as the size of the group increases. The validity of the Theorem is based on the hypothesis, that the answers provided by each individual are not random - but instead there is a more than 50% probability to be correct.
The Condorcet Jury Theorem might be criticized when we are dealing with humans, but machines can be programmed to work in such a manner.
It seems to me, that this is how cultural revolutions emerge: combining disciplines, which at first seem irrelevant to bring forth innovative ideas and technologies.
Probabilistic chips, can change the scenery of energy consumption, mobile technologies and microprocessor development; and all that by applying an almost 200 old theory to a completely different research field.
