Slashdot Mirror
New Hardware Needed For Future Computational Brain
schliz writes "Salk Institute director Terrence Sejnowski has called for more power-efficient, parallel computing architecture to support future robots that could keep up with the human brain. While human brains had 100 billion neurons and required only 20 Watts of energy, today's most powerful supercomputer, the 2.57 PFlop Chinese Tianhe-1A, requires four megawatts, and still has trouble with vision, motion, and 'common sense,' he said."
That most powerful supercomputer, I'd assume, has not been tuned to actually work like a brain would.
This is like an emulator. A lot of computational power is probably wasted on trying to translate biological functions into binary procedures. I think if they truly want to compare, they'll need to create an environment that is enhanced for the tasks we want it to process.
Nobody expects the human brain to compute integer and floating point stuff at the same efficiency either, right?
I think one mistake (besides the power requirements) that people make is to assume "if you build it it will work from the start", the human brain needs over ten year to develop even mediocre common sense and awareness of its surroundings. We should not be able to just build the hardware, install the software, flip a switch and then expect the machine to fully function the first year even. A learning period for the machine is to be expected (though it might be accelerated to some degree) if it is going to work like a human thinks.
Instead of trying to emulate the human brain, which at the moment is unattainable, we should concentrate on efficiency paradigms of smaller neural ensembles. Once we achieve efficiency we can scale. Why haven't we learned anything from the CPU industry? They didn't start from 19nm manufacture. Why should we?
We shouldn't hurry. AI comparable to a human person can be achieved, but it is still a long way until we reach it.
"Sum Ergo Cogito"
Each pint of beer contains 600 joules of energy, which can power your 20 watt brain for many hours, and give you trouble with vision, motion, and common sense.
The significant number is interconnect. In that area electronics is several orders of magnitude farther behind. Far enough that is seems doubtful something even remotely like the interconnect of a human brain can be reached artificially.
Side note: Comparing neurons and transistors, as is often done in the popular (but not very knowledgeable) press, is completely invalid as well. You need to compare neurons more to a micro-controller each.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Its interesting that you think epistemology actually plays a part for the flipping computer.
I could only agree if we are speaking of computer that is intending - by and within its design - to learn like, as well as act like us in a mature state. I agree this may be the most pure way for getting AI to resemble the human condition (for a lack of a better way to put it), but executing on this path is entirely a red herring.
I would say that trying to understand and emulate the learning process is 10 to 100 orders magnitude over the the effort of just getting the damn thing to work at a common, layman intellectual level.
We have no real understanding how we learn, empirically scientifically speaking - we are only beginning to understand this now. The understanding of this process changes rapidly and while we think we have momentum currently, more major unknowns exist. In fact, we don't know what we dont know at this point.
Its been debated as long as man has had the ability too, however... but even throughout the thousands of years of philosophical deep diving, it wasn't until the age of enlightenment that Kant finally got everyone on board for "Epistemology First" in our understanding of our world - we must first understand how we learn about this place, before we can debate the ontological status of the world around us and have any meaningful debate of its metaphysics. Theocratic or not, this rings true - and its only added more complexities to the struggle of what we know about ourselves.
And now, you want to build a robot to approach this condition.. Insanity. The effort is pure insanity and full of hubris. Lets work on simple tasks, and try to get those right, first. And how baout an honest look on who the fuck we are as emotional, sentient, chemically riding and wicked imperfect machines ourselves, before we attempt to perfect it in a model.
The only real saving grace is that this effort could actually be such a mirror for man kind, and accelerate our understanding of ourselves, if only slightly.
Eg build a hive mind.
Like 4chan?
-- Linux user #369862
This is a valid point. There is indeed a learning factor for the brain... at least some aspects of the brain.
Our brains are extremely inaccurate. Our perceptions are always relative and demonstrably imaginative. There is a lot more to what we think we see and know versus what we actually see and know.
The thing with computers as we currently use and design them is that they are dependant on accuracy. (I recall when DRAM was coming into existence... people were flipping out over the idea that this type of RAM needs constant refreshing to remain accurate and many doubted it was even possible.) Accuracy requires power. Our brains, instead, use an inaccurate and error-prone system with a LOT of built-in error checking and redundancy and even then is only generally accurate while using less power.
So far, it is even hard to imagine a processing system like our brains because we don't even think in the way our brains actually work. And to accomplish this, people will first have to begin to accept that brains do not work like a binary based digital device which is hard enough as it is.
Getting a little ahead of ourselves aren't we?
We're still not entirely sure of how a brain works. Oh sure, it's a neural network of some kind, but how do the neurons in a brain form meaningful connections with each other? How do they get their weightings of activation? etc.
Chances each neuron in the brain might be representable by a simple mathematical function with only a few terms. The way the neurons connect to each other might also be representable in a simplistic way. (btw. look up dynamic markov coding if you want to see a neat way a state can reproduce in a way that gives the newly created state meaningful input/output connections to other states).
So the problem isn't necessarily that our computers aren't powerful enough. The problem is that we still don't know how a brain works.
Most humans could never approach the capabilities of a common calculator.
Rocket Surgeon.
It's a software problem.
The architecture on which you run the software also determines quite a lot of what you can do and how the software is executed. You need a certain topology of the hardware, otherwise it is impossible to do certain tasks efficiently. There is a huge difference between a slow but massively interconnected network like the brain, and a sequential microprocessor running instructions one by one at high speed.
There was an article in Discover last year sometime describing the different techniques computer scientists were using to try to emulate/simulate a human brain. One of the more interesting is one that actually used simple software to create several thousand neurons, each able to communicate with thirty or so other neurons, and they made the pathways changeable.
Obviously I'm simplifying and paraphrasing a year old article here, but one of the most intriguing things about this one setup is not only that it apparently 'paid attention' when various objects were held in front of its cameras, but when the cameras and mics were shut off, the software neurons still showed waves of activity completely independent of outside input. To anthropomorphize a bit, it was almost like a sleep-state or beta waves in humans. I found that to be incredibly interesting.
"Who's brain did you emulate?"
"Uh, Abby someone..."
"Abbey who?"
"Abby Normal...."
Interview with Henry Markram This is the guy the article was about, but for the life of me I can't find the actual article where they describe the brain 'lighting up like a christmas tree', though I remember that exact phrase. Still, this describes his work pretty well. So might be worth a read.