We’d been through eight hours of lectures at the Singularity Summit, talks about the future of man-machine interfaces, anti-aging technology, human values in a post-human world.
We’d been through eight hours of lectures at the Singularity Summit, talks about the future of man-machine interfaces, anti-aging technology, human values in a post-human world. But it was at the end of a long and mind-taxing day that two scientists debated the fundamental question in artificial intelligence: Do we know enough about the brain to build machines that attempt to replicate it (or even parts of it?)
The skeptic was Dennis Bray, a Cambridge neuroscientist. He represented carbon-based intelligence, the kind that is carrying out extravagently complex tasks as you make sense of these words. For a half hour, he led us through the workings of a single cell, and discussed the mysteries that remain to be discovered. That’s one cell. And the human brain has 100 billion neurons, each of them making uncounteded and poorly understood connections with others. Even the connections have modulations. It’s a phenomenally complex network, and we’ve barely started to decode its workings. How, he argued, can we attempt to model machines on something we don’t understand?
On the other side was Terrence Sejnowski, who heads the computational biology lab at the Salk Institute. To the sound of 2001 A Space Odyssey he showed a computer simulation of the release of a neural transmitter. He agreed with Bray that the complexity was daunting, but said that with the exponential growth of computing, and the learning that accompanies it, scientists would be able to model the brain. The transmitter, he said, was an early step. …quot;We’re taking it one step at a time….quot; But he added that …quot;even if the models are incomplete, they’ll show us what’s missing. Then we’ll look for the missing pieces….quot;
Will they find the missing pieces in time for the Singularity? That’s the point, in about 2029, according to Ray Kurzweil, when computers should pass humans in intelligence. Well, if machines continue their march, they should increasingly help measure and model the workings of the brains that are building them. That’s the exponential factor Sejnowski refers to. But listening to Bray, it became clear to me that no matter how much complexity we unravel, we’ll always be confronted with more, much more.
The other question is whether the brain is the right model for computer. Early aviators studied birds. But it was a decidedly non-bird-like machine that finally led to the age of aviation. And the vessel that carried me from Newark to San Francisco two days ago was closer in its model of propulsion to an octopus than an eagle.
Heading back to the conference today. Just more thought about complexity. It’s not only cells that are complex, but every moment in time. (And each cell evolves through time. Your brain has changed since you started reading this post.) In his poem, 1964, Jorge Luis Borges wrote: …”Un instante cualquiera es mas profundo y diverso que el mar….quot; (A single moment is deeper and more diverse than the sea….”)
