Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

IBM announces advances toward a computer that works like a human brain

siliconvalley.com

Post reply on HN
Print view View original

11–20 of 25 posts

Re: IBM announces advances toward a computer that works like a human brain

#11

Post by glymor »

SlyShy wrote: post #2

I find this a fascinating area of research, however a prediction of my dad's always comes to mind. "A computer as intelligent as a human will take 18 years to 'program.'"

I know that's just supposed to be a piece of folky wisdom but it doesn't represent a very good understanding of how different AI is from our normal expectations.

It's not Data on Startrek. There's no reason for them to be walking around talking. You plug it directly into whatever you want it to understand: weather readings, big data sets, etc. You can 'program' it just as fast as it's processors allow it to consume. The analogy is how humans understand language or music although it's still far more direct.

Humans don't represent a upper bound on intelligence. The question isn't a computer as smart as a human, it's one several orders of magnitude faster or slower than one.

Re: IBM announces advances toward a computer that works like a human brain

#12

Post by nwatson »

JunkDNA wrote: post #6

I'm always suspicious of anything that people claim "works like the human X". In many cases (the brain especially) we have only the most rudimentary understanding of how things actually work. Our ability to create something that works the same way is therefore only as good as that rudimentary understanding. That said, it's clear they are trying to use this computer to understand more about brains. But whenever I read…

Many believe in 'emergent' phenomena where if you put enough components together, put them in the right environment, and stress them properly they will react in unexpected and coherent ways that while describable and maybe useful from an external perspective can't fully be analyzed because the interrelationships are too complex.

People were building bridges before they knew why the atoms in a beam of wood stuck together -- it probably wasn't even a matter they contemplated. They built complex bridgess before knowing fully how to analyze tensions and torques in such a structure. Likewise people will probably build useful biological wetware components and assemblies before knowing fully how they work.

Re: IBM announces advances toward a computer that works like a human brain

#13

Post by randallsquared »

There's a lower limit for how much processing power it will take to run a human-level intelligence. The highest that lower limit can be is at the level where we can simulate each neuron in a human brain; at that point, we can build an AI merely by simulation. It seems likely, though, that by the time we get to being able to simulate an entire human brain, we'll long have passed the actual lower limit for human-level intelligence, if we knew exactly how to build it. Assuming we survive, it'll be interesting to find out when we passed the lower limit with supercomputers, when we understand enough to figure it out.

Re: IBM announces advances toward a computer that works like a human brain

#14

Post by jpeterson »

randallsquared wrote: post #13

There's a lower limit for how much processing power it will take to run a human-level intelligence. The highest that lower limit can be is at the level where we can simulate each neuron in a human brain; at that point, we can build an AI merely by simulation. It seems likely, though, that by the time we get to being able to simulate an entire human brain, we'll long have passed the actual lower limit for human-level…

The highest that lower limit can be is at the level where we can simulate each neuron in a human brain; at that point, we can build an AI merely by simulation.

This assumes, of course, that we'll be able to observe every possible action of a neuron, and that these actions are all finitely computable.

Re: IBM announces advances toward a computer that works like a human brain

#16

Post by gloob »

ballpark wrote: post #15

Man is working so hard to learn and recreate the human body. However, many believe that the human body was created by chance and chaos.

I believe that hills were created by chaos and chance. Doesn't mean we're incapable of constructing ramps.

Re: IBM announces advances toward a computer that works like a human brain

#17

Post by ballpark »

gloob wrote: post #16
ballpark wrote: post #15

Man is working so hard to learn and recreate the human body. However, many believe that the human body was created by chance and chaos.

I believe that hills were created by chaos and chance. Doesn't mean we're incapable of constructing ramps.

I think it's great that we're learning and exploring the human body, and trying to mimic it in other areas. I'm really attempting to challenge the beliefs.

*edit - wording

Re: IBM announces advances toward a computer that works like a human brain

#18

Post by jimbokun »

nwatson wrote: post #9

quote: "Modha imagines a cognitive computer that could analyze a flood of constantly updated data from trading floors, banking institutions and even real estate markets around the world — sorting through the noise to dentify key trends and their consequences." ... and we'll have at the end a pseudo-human opinion that also will totally fail to see the housing bubble. It will also have opinions about the existence of G…

That quote along with

"A cognitive computer might also help soldiers analyze and react to chaotic events on a battlefield."

seems to be there for the benefit of people controlling the purse strings of the grant money. DARPA (or IARPA now) and potential backers from the financial world (although, do finance folks actually finance this kind of research?).

It may actually work at predicting financial trends, for a while. Then some trading floor will utilize that information to trade, and then their competitors will, then these machines will totally fail to see some other consequence of the complex trades they are making with each other.

Re: IBM announces advances toward a computer that works like a human brain

#19

Post by a-priori »

wglb wrote: post #10
JunkDNA wrote: post #6

I'm always suspicious of anything that people claim "works like the human X". In many cases (the brain especially) we have only the most rudimentary understanding of how things actually work. Our ability to create something that works the same way is therefore only as good as that rudimentary understanding. That said, it's clear they are trying to use this computer to understand more about brains. But whenever I read…

I think you have nailed it. I don't actually think we have a clue as to how the brain actually works. A while ago, we think it is like a telephone exchange, then we think it is like a computer, then we think it is like the cloud. The true advances in interesting computer results comes not from modeling the brain, but more from low-grade mimicking of the evolutionary process.

A while ago, we think it is like a telephone exchange, then we think it is like a computer, then we think it is like the cloud.

There's a big difference between a metaphor and a theory. No one actually thinks the brain works like a computer. It's just a useful metaphor for working memory and other psychological phenomena. Quite far removed from the neuroscience underlying it.

The true advances in interesting computer results comes not from modeling the brain, but more from low-grade mimicking of the evolutionary process.

What does this even mean?

Re: IBM announces advances toward a computer that works like a human brain

#20

Post by a-priori »

randallsquared wrote: post #13

There's a lower limit for how much processing power it will take to run a human-level intelligence. The highest that lower limit can be is at the level where we can simulate each neuron in a human brain; at that point, we can build an AI merely by simulation. It seems likely, though, that by the time we get to being able to simulate an entire human brain, we'll long have passed the actual lower limit for human-level…

Right now, it's possible to simulate one biologically-accurate neuron at a compartment-level (where you model a neuron by dividing it into many connected compartments), on one CPU core in real-time.

Here's the kicker: it's also been shown that these simulations can scale linearly with the number of processors (given sufficient interconnect bandwidth). So with N cores, you can simulate kN neurons for some constant k (and I think that k=1).

This implies that the limit is the number of CPU cores you throw at the problem. The human brain has around 100 billion neurons so, in theory, if you had a 100-billion core supercomputer, you could simulate the whole thing. That's a lot, but my point is that scaling a simulation is an engineering problem.

This doesn't say anything about the accuracy of the model; that depends on accurate characterisations of each part of the neuron, and of how they're arranged and connected. But the key point here is that these are low-level characteristics that can be individually examined, making whole-brain simulations a tractable problem.

Source: Djurfeldt et al. Brain-scale simulation of the neocortex on the IBM Blue Gene/L supercomputer. IBM JOURNAL OF RESEARCH AND DEVELOPMENT (2008)

Post reply on HN