DID YOU KNOW?

[jpvideo]

[madprof]

Interesting up until the "Computers in 2013 will have more computational power than a human brain" bit.

Having more transistors in a CPU than the brain has neurons doesn't imply greater computational power. I am pretty sure I could design a chip with 6 trillion transistors that didn't do much more than change some traffic lights. Although it might be a bit inefficient.

[markone]

Nov 23, 2009 22:34:57 GMT madprof said:

Interesting up until the "Computers in 2013 will have more computational power than a human brain" bit.

Having more transistors in a CPU than the brain has neurons doesn't imply greater computational power. I am pretty sure I could design a chip with 6 trillion transistors that didn't do much more than change some traffic lights. Although it might be a bit inefficient.

There's been some interesting work done on calculating the actual processing power of the brain (rather than actual transistors) at Carnegie Mellon University robotics institute, they estimate the human brain has the equivalent processing power of about 100 Teraflops.

IBMs ASCI Purple supercomputer is supposed to be that powerful, so in theory a computer with the processing power of the human brain is possible today.

But the thing is the processing algorithms associated with strong AI just don't exist yet. Weak AI - which is artificial intelligence designed to solve a single problem, or manage a particular task, is something that researchers can get a handle on (Deep Blue is a good example of Weak AI; with the ability to outthink a chess grand master, but that is it's ONLY trick) Strong AI would have a general capability to learn many tasks. But that is in it's infancy right now, and some researchers are just beginning to suspect that rather than being just over the horizon, strong AI might be decades away at least.

A computer able to pass the Turing test is unlikely to be around the corner. But here's some food for thought. A software algorithm based on natural selection and evolution was tasked with a trial and error based design of a simple electronic circuit (a device called a flip-flop oscillator - something that year one electronics students build) The program was successful and created a fully functional circuit. However when the researchers looked at it's solution there was a component that clearly looked redundant and appeared to take no part in the functioning of the circuit. When they removed it the circuit stopped working. But they had no idea why!

Now imagine a weak AI given the single task of developing a better version of itself. How many iterations might it take before there was a fully functional system that worked fine, but we had NO IDEA how it did it.

Fascinating stuff. Even if I don't subscribe to Ray Kursweill's view that the Technological Singularity is just around the corner and that machines with minds immeasurably greater than ours will look down on us like we do nematode worms.

[Shreddie]

I think that the biggest hurdle to overcome with AI will be giving the machine curiosity. Curiosity leads directly to learning afterall.

[madprof]

Don't agree with you there. Newborn babies don't have personalities as we do. Their brains aren't wired up in the same way. They aren't "curious" but they do learn.

If I've misrepresented your idea of "curiosity" then please do correct me.

[psionic11]

I think a more accurate way to put it is this -- we may be able to program various algorithms which can calculate math, manage internal physical states, manage memory resources, efficiently mine knowledge bases, and even construct passable poetry or music using pre-defined rules, but a large hurdle will be coding for "observing a gap in knowledge that could be bridged by a relevant, spontaneous and feasible search for more data and patterns that could then be integrated into the greater knowledge-base that would help further survival/growth"...

That's a fancy way to explain "curiosity."

[Shreddie]

Dec 8, 2009 23:27:23 GMT madprof said:

If I've misrepresented your idea of "curiosity" then please do correct me.

Ok!

They aren't "curious" but they do learn.

Curiosity works on many levels. Newborns may not be curious in quite the same way that older children are because they are not yet sufficiently developed. However, they do have the most important and most basic building block of 'curiosity'... That's paying attention to the world around them, and especially those parts of it they find interesting. Babies eyesight is poor for the first few weeks, they can only focus on objects about 6-12 inches away. It's thought that this is because it helps them to be less distracted and only allows them to focus on things that matter such as the mothers/fathers faces when held and the breast when feeding. Their hearing however is fully functional, even at just a few minutes old, newborns will react to sounds they find interesting or are used to. Such as their mothers voice which they recognise from what they have heard in the womb.

Over time, that paying attention will lead on to thinking about what they see, hear and experience, thinking about that will lead to questioning it and that is basic curiosity.

As for the "Their brains aren't wired up in the same way" bit... That's actually very interesting. It's now thought that when you are born, everything is wired up in one massive mess. Over time, the less used connections die off and the most used connections are reinforced leaving a more organised brain. It's believed that some people (such as myself) who have conditions like synesthesia still have some connections that the rest of the population have lost. It's thought that it's because we learn through association (which is a tried and trusted memory trick btw) so those associations (connections) which work remain while those which don't do not. As for those who's connections are not wiped out, some people have letter>colour synesthesia where all letters they see are coloured the same way that the fridge magnet letters they played with as a child were. Others who perhaps had a sister with a record deck that had 'disco lights' which were triggered by music (like me) have audio>colour synesthesia. Others may taste spoken words, others may smell colours. Synesthesia is different for everyone who has it but everyone has it to some degree.

For example, I vaguely remember a test that has been quite popular so I'll try that on you... Though I can't remember the exact names usually used.

Imagine that you have two shapes, one is an irregular but rounded shape, the other is an irregular but spiky shape. One is called Keekee and the other Booboo. Which one is which?

I'll tell you the usual results to this in a day or five.