I can feel it in the air, so thick I can taste it. Can you? It's the we're-going-to-build-an-artificial-brain-at-any-moment feeling. It's exuded into the atmosphere from news media plumes ("IBM Aims to Build Artificial Human Brain Within 10 Years") and science-fiction movie fountains...and also from science research itself, including projects like Blue Brain and IBM's SyNAPSE. For example, here's a recent press release about the latter:
Today, IBM (NYSE: IBM) researchers unveiled a new generation of experimental computer chips designed to emulate the brain's abilities for perception, action and cognition.
Now, I'm as romantic as the next scientist (as evidence, see my earlier post on science monk Carl Sagan), but even I carry around a jug of cold water for cases like this. Here are four flavors of chilled water to help clear the palate.
The Worm in the Pass
In the story about the Spartans at the Battle of Thermopylae, 300 soldiers prevent a million-man army from making their way through a narrow mountain pass. In neuroscience it is the 300 neurons of the roundworm C. elegans that stand in the way of our understanding the huge collections of neurons found in our or any mammal's brain.
This little roundworm is the most studied multicellular organism this side of Alpha Centauri -- we know how its 300 neurons are interconnected, and how they link up to the thousand or so cells of its body. And yet ... even with our God's-eye-view of this meager creature, we're not able to make much sense of its "brain."
So, tell me where I'm being hasty, but shouldn't this give us pause in leaping beyond a mere 300 neurons all the way to 300 million or 300 billion?
As they say, 300 is a tragedy; 300 billion is a statistic.
Big-Brained Dummies
About that massive Persian army: it didn't appear to display the collective intelligence one might expect for its size.
Well, as it turns out, that's a concern that applies to animal brains as well, which can vary in size by more than a hundred-fold -- in mass, number of neurons, number of synapses, take your pick -- and yet not be any smarter. Brains get their size not primarily because of the intelligence they're carrying, but because of the size of the body they're dragging.
I've termed this the "big embarrassment of neuroscience," and the embarrassment is that we currently have no good explanation for why bigger bodies have bigger brains.
If we can't explain what a hundred times larger brain does for its user, then we should moderate our confidence in any attempt we might have for building a brain of our own.
Blurry Joints
The computer on which you're reading this is built from digital circuits, electronic mechanisms built from gates called AND, OR, NOT and so on. These gates, in turn, are built with transistors and other parts. Computers built from digital circuits built from logic gates built from transistors. You get the idea. It is only because computers are built with "sharp joints" like these that we can make sense of them.
But not all machines have nice, sharp, distinguishable levels like this, and when they don't, the very notion of "gate" loses its meaning, and our ability to wrap our heads around the machine's workings can quickly deteriorate.
In fact, when scientists create simulations that include digital circuits evolving on their own -- and include the messy voltage dynamics of the transistors and other lowest-level components -- what they get are inelegant "gremlin" circuits whose behavior is determined by incidental properties of the way transistors implement gates. The resultant circuits have blurry joints -- i.e., the distinction between one level of explanation and the next is hazy -- so hazy that it is not quite meaningful to say there are logic gates any longer. Even small circuits built, or evolved, in this way are nearly indecipherable.
Are brains like the logical, predictable computers sitting on our desks, with sharply delineated levels of description? At first glance they might seem to be: cortical areas, columns, microcolumns, neurons, synapses, and so on, ending with the genome.
Or, are brains like those digital circuits allowed to evolve on their own, and which pay no mind to whether or not the nakedest ape can comprehend the result? Might the brain's joints be blurry, with each lower level reaching up to infect the next? If this were the case, then in putting together an artificial brain we don't have the luxury of just building at one level and ignoring the complexity in levels below it.
Just as evolution leads to digital circuits that aren't comprehensible in terms of logic gates -- one has to go to the transistor level to crack them -- evolution probably led to neural circuits that aren't comprehensible in terms of neurons. It may be that, to understand the neuronal machinery, we have no choice but to go below the neuron. Perhaps all the way down.
...in which case I'd recommend looking for other ways forward besides trying to build what would amount to the largest gremlin circuit in the known universe.
Instincts
It would be grand if brains could enter the world as tabula rasa and, during their lifetime, learn everything they need to know.
Grand, at least, if you're hoping to build one yourself. Why? Because then you could put together an artificial brain having the general structural properties of real brains and equipped with a general purpose learning algorithm, and let it loose upon the world. Off it'd go, evincing the brilliance you were hoping for.
That's convenient for the builder of an artificial brain, but not so convenient for the brain itself, artificial or otherwise. Animal brains don't enter the world as blank slates. And they wouldn't want to. They benefit from the "learning" the countless generations of selection among their ancestors accumulated. Real brains are instilled with instincts. Not simple reflexes, but special learning algorithms designed to very quickly learn the right sorts of things given that the animal is in the right sort of habitat. We're filled with functions, or evolved capabilities, about which we're still mostly unaware.
To flesh them out we'll have to understand the mind's natural habitat, and how the mind plugs into it. I've called the set of all these functions or powers of the brain the "teleome" (a name that emphasizes the unabashed teleology that's required to truly make sense of the brain, and is simultaneously designed to razz the "-ome" buzzwords like 'genome' and 'connectome').
If real brains are teeming with instincts, then artificial brains also want to be; why be given the demanding task of doing it all in one generation when it can be stuffed from the get-go with wisdom of the ancients?
And now one can see the problem for the artificial brain builder. Getting the general brain properties isn't enough. Instead, the builder is saddled with the onerous task of packing the brain with a mountain of instincts (something that will require many generations of future scientists to unpack, as they struggle to build the teleome), and somehow managing to encode all that wisdom in the fine structure of the brain's organization.
The Good News
Maybe I'm a buzz kill. But I prefer to say that it's important to kill the bad buzz, for it obscures all the justified buzz that's ahead of us in neuroscience and artificial intelligence. And there's a lot. Building artificial brains may be a part of our future -- though I'm not convinced -- but for the foreseeable, century-scale future, I see only fizzle.
~~
Mark Changizi is an evolutionary neurobiologist, and Director of Human Cognition at 2AI Labs. He is the author of The Brain from 25000 Feet, The Vision Revolution, and his newest book, Harnessed: How Language and Music Mimicked Nature and Transformed Ape to Man. This piece first appeared Nov 16, 2011, at Discover Magazine.
Follow Mark Changizi, Ph.D. on Twitter: www.twitter.com/markchangizi
It's by Carver Mead, one of the greats of the whole microelectronics revolution.
He uses analogs, literally analog computers right down to the transistor matching a dendrite and variable analog resistors the weights. I don't know if he has done the 300 neuron worm, but I bet he could.
He is not just a great scientist, but a great writer who explains these incredibly complex systems withe simple beauty.
His analogy of a dam to a MOSFET is not to be missed.
As to the need for the physical body, I tend to agree. you need the whole closed loop sensory to action system for the "mind" to learn.
That's why most of Mead's circuits are audio or visual processing circuits.
I think it will start, as it already has, with circuit being connected to our brains. First to fix problems, like the artificial cochlea, but then cognitive problems, we already have the circuit that can connect to 1000s of neurons per chip.
After all, how many analog computers have you seen in your life?
You really miss the bigger point, who want artificial neural networks and for what?
The military uses them for weapons tracking and recognition in particular. They have great speed, and complex abilities but fairly low resolution about 8 bits. so as soon as the digital circuits are fast and dense enough they replace the analog sections. But at the limits of speed and power, it's always analog.
How many analog computers have I seen? Every singe computer, phone, and nearly all electronic devices with any kind of sensing or radio involve rather large and complex analog computational units in the front ends.
The linear scaling between animal size and brain size is surely just due to the constant density of microscopic nerves in mammal tissue - if the signals all end somewhere then they need a destination. Whales and dolphins are outliers because seeing by sound in water is a challenge.
As a student studying bioengineering and premed for neurology I admit that the author has a little more experience than me. I've always found that someone who carries around a "bucket of cold water" is just looking for an excuse of how it can't be done.
Computers (without being designed like a brain) are getting closer and closer to the intelligence capabilities of the human mind and are already capable of much more raw information processing today. What do we do when we can't figure something out? We give it to someone who is able to handle the problem... and already computers are doing this for us. Watch the curve of the graph, we will get there.
Do i think this is 2, 4, or 10 years ahead? Maybe, and unlikely.... but 20.... reasonably certain.
I doubt computational power correlates to self awareness. Who knows though.
One can, of course, argue that the structure of our operating systems doesn't allow for any "evolution" of a distributed intelligence. One could start an interesting experiment... if millions of computer owners would allow their machines to be taken over by a self-evolving, distributed AI algorithm during idle cycles... it might actually lead to some interesting results.
However, is a general purpose brain what we're after? Swarm technology took the path that intelligence is emergent from a few simply parts left to act chaotically to the point that some useful function appears. And it seems to succeed.
And so what if we don't actually "build" a human brain (and in some sense, each and every mother can do that with childbirth), if a Boeing 787 a gigantic bird with feathers and flapping wings? And would we ever want it to be?
Human intelligence is, so far, the only known example of fuzzy reasoning that is MUCH more successful than domain specific information processing, and it is only successful for enormously complicated problems... like performing science, engineering and creating art that appeals to humans. Even we suck on pretty much everything else.
Glial cells in the human brain seem to be on the same order as neurons. If we add their function to the complexity, we are still on the same order of magnitude. Of course, if size alone would matter, the internet would have become conscious about twenty years ago... but in nature size without structure equals thermodynamics, not intelligence and consciousness.
Perception, thinking, response are growth processes. At the moment, we do not have an AI construction which can grow, pare unneeded connections and create new connections. We have the ability to mimick certain behaviors, but we really don't yet how to actually produced the behavior, or to give the AI the same perception of stimulus and response.
I agree with the author that true AI is a long way off. We may well produce incredibly complex programming. We might perfect self-correcting algorithms. But hardware intelligence may take a while.
Nothing stops you from programming one on the computer you wrote this comment on. It's a Turing machine, so it can do all the things you want. Synapses are pointers in unordered lists. Probably wouldn't take much more than a couple of hundred lines of code, chemical modulation included.
The problem is... that code wouldn't do anything, because we don't know how to set the initial connections, which are not random. OTOH, we know that the total information content in the human genome is only on the order of 1GByte. So, even under the worst circumstances, the plan for the human brain fits into a smart phone several times.
The machine to emulate a brain is not that far off, either. A human brain has about 1e15 synapses, firing at 1-2kHz, each, which can be reduced to probably 1e19 binary operations per second.
A single modern FPU does about 1e14 binary operations per second. So that's no more than five orders of magnitude less. A thousand of these can be put into a normal server rack, so that's a hundred racks. That's it... a medium sized custom supercomputer.
But it wouldn't matter if you built it... you couldn't make it do anything without knowing, at the very least, some of the principle circuits that the brain uses to create what we call "us".
Also, you seem to dismiss or simply miss that we are on an exponential growth curve with regards to technology, anybody attempting to make ANY prediction 88 years into the future, and most especially those trying to say what will NOT be possible, are talking out of their rear.
We are very far away from that goal, too, but at least it is reasonably achievable within the next 100 years.
Recall a mere hundred years ago, despite Newton, gravity was still intangible, not well understood, until Einstein visualized and simplified the vision with mathematics and from help of determined astronomers.
While it seems an impossible task, so does trying to see beyond Singularity, Dark matter, or beyond the edge of the Universe (Multiverse Theory...brain), and Big Bang. Theoretical physicists are undaunted. Likewise, we're curious about matter/energy connection forming consciousness.
And where there's energy...physicists are always on the scent/trail, no matter how difficult. So, perhaps, you may reconsider and envision another young Einstein will unlock the mystery, describe it in detail, and win a Nobel prize.
A few reproducible experiments are available already (about the same as Dark matter), so we know there's something on the periphery, and it's worth pursuing, especially, physicists.
Gravity was in the domain of physics 400 years ago and is in the domain of physics today. It will remain in the domain of physics by definition.
Information processing in the brain wasn't in the domain of physics 400 years ago, is not in the domain of physics today, by definition. You can, of course, apply the tool set of theoretical physics to the problem and many have. That doesn't make the problem a physics problem. It might make some theoretical physicists smarter than most of the people in the field, though.
"SingularitÂy, Dark matter, or beyond the edge of the Universe"
Neither of which have anything to do with any of this.
"And where there's energy...pÂhysicists are always on the scent/traiÂl, no matter how difficult."
Physicists like to play with everything. But they still don't call their pet projects physics.
"So, perhaps ... win a Nobel prize."
There is no Nobel for biology. This could be a Nobel for physiology or medicine. It certainly won't be one for physics.
"A few reproducibÂle experimentÂs are available already (about the same as Dark matter), ...."
All dark matter experiments, so far, are either astronomical observations or particle physics null experiments. All brain related experiments, so far, are essentially psychology studies with functional brain imaging or animal models. I fail to see the connection, theoretically, experimentally, or even on the level of methodology.
- Mark Changizi (PhD in computer science, former computer science professor at University of Ireland, Cork).
In any case... if you want, I can teach you how to build a pretty good neuron analogy with a dozen transistors. Bipolar or CMOS, I don't care. It's pretty easy. Not that I have to... there must be dozens on papers on the topic. Ooops... I just found one on google!
Of course... our aerospace engineers are also not building planes that look like giant eagles, with plastic feathers... so it would be a completely useless intellectual exercise. Nature didn't use neurons because it wanted to. It used neurons because that's what was within the reach of evolution.
"It seems to me that the goal is to emulate nature as much as possible using advanced hardware and software."
Why? Bridges do not emulate stones in a creek. Buildings do not emulate caves. Cars do not emulate legs and planes do not emulate birds. Why would artificial intelligence emulate brains?
"To the extent that we can re-create the neuron-basÂed architectuÂre of the human brain, we have a chance to create human-like intelligenÂce."
All it takes to create human intelligence is a woman and a man. We have already too much of it. We can't even properly educate the majority of human brains we have. Why create more artificially?
"But I believe that intelligenÂce is an emergent property."
Functional brains are not. See deer. They are born, they get up in an hour and they get going. Cats and dogs, both of which show signs of intelligence, have rather short learning periods. Humans are different, but only to an extent. We have a much larger capacity to order information hierarchically, but that doesn't "emerge" in any one individual by itself. It is a matter of learning advanced structures that were developed by our predecessors... like science.
It's almost as though our neurons allow us to tap into something much larger, an energy outside of ourselves, an energy all around us, something obviously not yet perceptible in the current scientific paradigm, nevertheless, associated with the light spectrum.
When brain size is irrelevant, seems neurons (regardless of quantity) are possible, likely, receptors or conduits to something larger and outside of ourselves; like energy all around us.
You know, it's a strange thing, Cosmologists reject Astrology, but nothing seems to describe or predict human behavior so well, as individuals. Birth charts are strikingly descriptive. It's as if, when we're born, there are undetectable planetary gravitational forces or undetectable energy, which initiate triggers within each of us, affecting our personality, providing a basic framework or essence of who we are, guiding the way we perceive and react to the world around us...despite free will.
Like dark matter, energy or consciousness is intriguing and very poorly understood.
Long-ish predictions are always risky.
All the while, time marches on. The mountain of instincts that future scientists start unpacking tomorrow, say, can theoretically be completed at some point, but then the poor fools have MORE unpacking to do just to catch up from the last end point. To ignore that critical task is de-evolution, a job best left to Republicans.
Great article, Mark! Thank you.
Religion can be dangerous, religion disguised as science even more so.
Thank you for this article.