(These are excerpts from my book "Intelligence is not Artificial")
The Next Breakthrough: Understanding the Brain
Looking back, one can see that progress in neural networks came almost
simultaneously from physics (Hopfield discovered recurrent neural networks
studying the annealing process) and from neuroscience (Fukuyama discovered convolutions studying the cat's visual system).
These two insights led to a rediscovery and adaptation of the old mathematics of optimization and control methods.
I suspect that this will continue to be the case in the future.
But the physics is mostly old, whereas the neuroscience is new.
So most of the progress in A.I. is likely to come from neuroscience.
National and international research programs have a bad reputation among people
who don't like to pay taxes, but those programs have accounted for immense
progress over the last century.
Franklin Roosevelt's Manhattan Project (1941) to build the atomic bomb,
John Kennedy's Apollo Program (1963) to send a man to the Moon,
Richard Nixon's Arpanet (1969) to create a national network of computers,
and George H. Bush's Human Genome Project (1990) to decode the human genome
have delivered byproducts for countless disciplines (and millions of well-paid jobs for the tax-payers).
Barack Obama's BRAIN Initiative (2013), as well as the European Union's Human Brain Project (2013), have the same potential. (Believe it or not, BRAIN stands for Brain Research through Advancing Innovative Neurotechnologies). The BRAIN initiative contained the Machine Intelligence from Cortical Networks (MICrONS) project to reverse-engineer the brain, conceived by Jacob Vogelstein and led by David Cox at Harvard, by Tai Sing Lee at Carnegie Mellon University and by Andreas Tolias at the Baylor College of Medicine. The bad news is that, according to the OpenWorm project, we are not even simulated the Caenorhabditis Elegans worm yet. Long way to go.
Understanding the brain, however, may or may not be equivalent to understanding "us", as philosophers never tire of discussing: when we find out everything
about my brain, when we design the great neural network and write the great equations that fully describe my brain, will we find out everything about me? Is consciousness just a mathematical formula?
Having been trained as a mathematician in theoretical Physics,
i find an intriguing parallel with the dilemma faced by physicists.
Richard Feynman said "If all mathematics disappeared today, physics would be set back exactly one week". But Eugene Wigner was puzzled by how well mathematics describes reality, i.e. by (quote) "the unreasonable effectiveness of mathematics in the natural sciences."
In the paper "Why Does Deep And Cheap Learning Work So Well?" (2016) Henry Lin, a physicist at Harvard University, and Max Tegmark, a mathematician at MIT, advanced the hypothesis that multilayer neural networks may have something profound in common with the nature of our universe.
Niels Bohr was fond of the principle of complementarity, that there exist dual formulations of reality, such as the the particle and wave aspects of physical systems, and that you have to choose one or the other but are never allowed to mix them (a fact that had just led Werner Heisenberg to his famous principle of indeterminacy). Later in life Bohr expanded the principle of complementarity to deal with philosophical topics: the mind-body problem, vitalism, yin and yang, etc.
I suspect that Bohr would have happily accepted as both one and many the brain as purely computational mathematics and the brain as a conscious being.
Is neuroscience the final answer or just another question?
Back to the Table of Contents
Purchase "Intelligence is not Artificial")