Miscellaneous > The Lounge
Computer simulation of the cat brain scam
Aloone_Jonez:
This may look cool.
http://spectrum.ieee.org/computing/hardware/ibm-unveils-a-new-brain-simulator
Then Google a bit and you soon realise it's total bollocks.
http://www.eweek.com/c/a/IT-Infrastructure/Rival-Scientist-Calls-IBM-Cat-Brain-Simulation-a-Scam-384096/
Of course the model doesn't think like a cat which is still more intelligent than a computer. Put it this way, if a tiny bee brain with under 1 million neurons beats a supercomputer at complex travelling salesman problem, the machine has nowhere the intelligence level of a cat which has around 300 million neurons.
http://www.cbsnews.com/8301-501465_162-20020587-501465.html
People go around spouting crap like human brain simulation will be possible by 2025 but they forget that computers may be powerful enough by then but without an accurate model it won't be possible. I'm pretty sure that computers are already powerful enough to simulate a bee brain but there obviously isn't a good enough model otherwise they would've done it by now.
http://en.wikipedia.org/wiki/Whole_brain_emulation
Does anyone think that AI will eventually happen? It's theoretically possible of course but there could be some fundamental law with means that humans aren't intelligent enough to make a machine with equal intelligence. Still, even if the smartest scientist can make a machine as bright as someone who's slightly retarded it would be great.
I think what some people don't understand is that brain size isn't the most important factor in intelligence but EQ (brain mas when body size (for example, a 10 gram crow brain is more intelligent than a 500 gram cow brain) is taken into account) and that brains are not computers. A different approach than brute force simulation is required.
IBM would've been better off trying to simulate the brain of a snail accurately, it certainly would've befitted neuroscience more.
piratePenguin:
I'm really interested in studying AI. I'm about 40% sure I'll aim to do a masters in some AI course after my degree.
I have no ideas about making AI that is as intelligent as a human, but I never read about human brain modelling before (I'll read these articles soon, sounds interesting but also sounds like huge hype). I've learned about artificial neural networks using back-propagation and a couple of simple (and common) methods so that a program can recognize features in data after being trained. I think there'll be HUGE advancements to be made year-on-year in the near future, hence my interest (also it ties maths and computers together), but I don't anticipate intelligence that can match humans.
The thing is when you add enough little intelligence to a machine, it can hammer a human at tasks it's trained to do.
Try and manually search a billion documents for keywords.
Computers are good at a lot of things, and in 5 years they'll be good at more things.
Then again, AI voice recognition is still pretty poor, and not in the least comparable to human beings after many decades. If there are similar problems bringing along other recognition systems it'll take something special to really bring AI forward. It doesn't turn me off, the subject definitely deserves growing attention.
davidnix71:
I've taken in a stray cat that came by begging food. She's smarter than any computer I've seen. It took her about two days to figure out my work routine. I put out a litter box and she used it without being told. She keeps herself clean enough that she can sleep on my bed without giving me fleas, either.
She's shown me her claws twice, but didn't draw blood either time. She sharpened her claws on my jeans without scratching me and groomed herself from my shoulders without falling off or scratching me.
She is insecure about food and water after losing a fight with something. She came to me with a scratched up face and a piece of an ear missing. I keep her indoors to keep the neighborhood racoons from steaing her food.
Even the Chinese don't have a computer fast enough to think like a cat.
Aloone_Jonez:
David,
Yes, cat's are great manipulators. Apparently they have a special meow call which is a similar pitch to a baby crying and they only use it for humans, not cats which implies they've figured out how to appeal to our paternal instincts.
piratePenguin,
Have you considered studying some biology? Would you consider doing a module on neuroscience?
I'm actually pretty amazed at how slowly the science of AI has progressed despite the massive improvements in computing power. Even though computers are capable of processing huge amounts of information, even more so than neural networks, they still don't do it intelligently.
I've not done much study into AI myself. I gather that it's progressed so slowly is due to computational complexity, you know to see a small improvement with some problems, computing power needs to increase exponentially which I believe is the case with AI, the bee brain beating the supercomputer at the travelling salesman problem is a classic example. But it's obviously not so complex, it's just the machines that the way that computers work that make it hard. We're definitely using the wrong tool to try to solve the problem but I admit I'm not smart enough to tell you what tool we need too solve it.
piratePenguin:
Hmm I've still not read much about bees but as I understand much of their impressive abilities are due to pheromones i.e. smells they can release to manipulate other bees. http://en.wikipedia.org/wiki/Honey_bee_pheromones
Take a look at this from Wikipedia, about ants:
--- Quote ---Ants communicate with each other using pheromones.[59] These chemical signals are more developed in ants than in other hymenopteran groups. Like other insects, ants perceive smells with their long, thin and mobile antennae. The paired antennae provide information about the direction and intensity of scents. Since most ants live on the ground, they use the soil surface to leave pheromone trails that can be followed by other ants. In species that forage in groups, a forager that finds food marks a trail on the way back to the colony; this trail is followed by other ants, these ants then reinforce the trail when they head back with food to the colony. When the food source is exhausted, no new trails are marked by returning ants and the scent slowly dissipates. This behaviour helps ants deal with changes in their environment. For instance, when an established path to a food source is blocked by an obstacle, the foragers leave the path to explore new routes. If an ant is successful, it leaves a new trail marking the shortest route on its return. Successful trails are followed by more ants, reinforcing better routes and gradually finding the best path.[60]
--- End quote ---
http://en.wikipedia.org/wiki/Ant#Communication
It's as clear that much of their "intelligence" isn't so much in their brain, it's in the way they evolved (natural selection - every ant that EVER thought they could make it on their own died a horrible death and didn't plant their seed, and thus ants need to be social, ...), and their brain only needs to follow the right instincts.
This is stuff that does fascinate me, I know very little but I think I will read an evolutionary biology book sometime soon. Studying neuroscience or some deep biology, I could probably do a module in that but probably not a year long masters or something. edit: I don't get any choices like that in my degree, most degrees I know are like mine. Can a maths student pick a biology module in colleges outside Ireland?
There is an technique for solving problems that works somewhat like natural selection, it's called genetic algorithms, it was interesting for me to look at. http://en.wikipedia.org/wiki/Genetic_algorithms
Navigation
[0] Message Index
[#] Next page
Go to full version