Friday, November 23, 2012

Gary Marcus on brain simulation — The Brain In The Machine

Half a trillion neurons, a hundred trillion synapses. I.B.M. has just announced the world’s grandest simulation of a brain, all running on a collection of ninety-six of the world’s fastest computers. The project is code-named Compass, and its initial goal is to simulate the brain of the macaque monkey (commonly used in laboratory studies of neuroscience). In sheer scale, it’s far more ambitious than anything previously attempted, and it actually has almost ten times as many neurons as a human brain. Science News Daily called it a “cognitive milestone,” and Popular Science said that I.B.M.’s “cognitive computing program… just hit a major high.” Are full-scale simulations of human brains imminent, as some media accounts seem to suggest?
Ray Kurweil is betting on a successful Turing test (by 2029), but Gary Marcus explains why the hurdles are high.

The New Yorker — News Desk
The Brain In The Machine
Gary Marcus
(h/t Noah Smith via Twitter)


Clonal said...

I think they are about 1/10^10 there - see Quantum computation in brain microtubules: The Penrose-Hameroff "Orch OR" model of consciousness

Classical microtubule automata switching in the nanosecond scale offer a potentially huge increase in the brain's computational capacity. Conventional approaches focus on synaptic switching (roughly 10^11 brain neurons, 10^3 synapses/neuron, switching in the millisecond range of 10^3 operations per second) and predict about 10^17 bit states per second for a human brain (e.g. Moravec, 1985). However as biological cells typically each contain approximately 10^7 tubulins (Yu and Bass, 1994), nanosecond switching in microtubule automata predicts roughly 10^16 operations per second, per neuron. This capacity could account for the adaptive behaviors of single cell organisms like paramecium, for example, who elegantly swim, avoid obstacles, and find food and mates without benefit of a nervous system or synapses. As the human brain contains about 10^11 neurons, nanosecond microtubule automata offer about 10^27 brain operations per second.

Matt Franko said...

"It’s sort of like having the biggest set of Lego blocks in town without a clue of what to make out of them. The real art is not in buying the Legos but in knowing how to put them together. Until we have a deeper understanding of the brain, giant arrays of idealized neurons will tell us less than we might have hoped. Simply simulating individual neurons without knowing more about how the brain works at the circuit level is like throwing Legos in a pile and hoping that they create a castle; what we really need are directions for creating the castle,"

this is interesting because key with mathematics its not that one knows how to do computations, it is that one knows how to apply it...

and many people would say that if you throw legos at a table enough times, a castle would eventually result "mathematically"...


Paulo Garrido said...

The question is that with the actually assumed physical models of computing systems no attempt is made to really consider consciousness. For consciousness I take here the subjective experiences of feeling, perceiving and acting.

Modern day physical computers are built and operated according to the Turing and contemporaries Wiener and von Neumann concept of computers as networks of electrical currents in wires. Such networks are able to exhibit intelligent behavior even on orders of magnitude greater than the exhibited by conscious beings on some dimensions, yet that does not the answer the initial question posed by the concept: how is consciousness to be correlated to electrical currents?

An analysis of the question reveals that there is no way one can find a connection between an electrical current in a wire and perception. If perception is to be correlated to a physical phenomenon it must be to a physical phenomenon more complex that ions moving in a medium.

The answer proposed by quantum consciousness models is that such correlate can be quantum electronic phenomena in the cytoskeleton system enhanced in the nervous system and brains – the microtubules system, gravitation affected in the Penrose-Hameroff model.

Our understanding of our perception flows as delicate quantum phenomena in living beings opens us great perspectives for rearranging our experiences and sense of ourselves, both as humans and living beings.

Only if we would be able to create beings, say androids, with in-built quantum computers could the question of creating beings conscious as us be realistically put. The question still goes to numbers: is it necessary a minimum number of coordinated bits to make up for perceiving as myself?

Current laboratory quantum computers are made with 2-4 bits at most. Assuming that in living matter quantum bits can exist in tubulins these gives numbers in the range of 10^18 or 10^6 times 10^12 or one million teraquantumbits (10^6) Tqb for a typical human brain.

An important consequence of quantum consciousness and the tubulin model is that were we trying to create quantum computers the best way to proceed would be… to produce living beings.

Which gots us into the more important consequence. All along living beings tubulins are encountered. If tubulins are the quantum bits of the quantum physical computers we are then some form of feeling, perceiving and acting as a subjective experience is going in virtually any living being.