Wait. Rik Mayall died?!?

No one dies permanently, the singularity will bring them all back!

Nice find. For some reason this picture tells me the most:

Graph without units on the axes are often a sign that the person who made the graph is doing nothing more than making crap up.

Puppy deals with the "singularity"

c'mon man

Good point. You need to read the text for getting more information and can partly blame me for taking out the picture from its context. The following pictures have units, from article one. It's still a matter of opinion how well you can make the raw capacity into performance, and probably to a lesser extent, if the raw capacity itself will grow approximately this way:



An the dramatic




Here is another picture, from the second article, which does have a scale



AGI (AI that’s at least as intellectually capable as a human, across the board)
ASI (AI that’s way smarter than any human, across the board)

We probably have the computing power to do it right now. The problem is computing power doesn't tell us how to do it.


PairTheBoard

Parallel computing may be one of the answers, but that's maybe not easily achieved before the advent of usable quantum computers. But with enough raw computing power you can compensate for that. Say, instead of having one million things going on simultaneously you can make a machine a million times faster. Not exactly the same thing, and if that's not enough, you can make it a trillion times faster and you will get there eventually, I figure.

Raw capacity isn't an issue. If you don't count the parts of the brain that are doing non-figuring-out-important-things, computers surpassed us quite a while back in raw capacity.

You are capable of keeping a maximum of 7 +/- 2 things in mind at a time, take forever to learn things, calculate far slower than even a stupid calculator, have horrible recall, etc.

With raw capacity I mean computational power inside the brain/computer, not the output. Raw capacity is needed at this stage for getting to the parts that are doing the "non-figuring-out-important things", as emotions, humor, good speach production, pain sensation etc. With the capacity of the brain of a mouse this can't be expected.

On the other hand, I predict there will be a sideline of research, a bit exclusive, where guys are looking at how much raw capacity is needed for passing the Turing test in multitudes of settings. It may be it will eventually be possible with a hundredth or a tousandth of the capacity our brains have.

It still depends on what you put it to work doing. If you tell it to search for prime numbers you can get it to run a trillion to the trillionth power faster if you want. It will find a lot of prime numbers but it won't be intelligent.

This "computing power" argument would make sense if they could point to applications we have now which they think require only more computing power to become something intelligent. They can't point to any such applications nor even to any ideas for such applications. Until they can we might as well be cave men with fancy hardware.


PairTheBoard

PTB,

We are cavemen with fancy hardware. You should pay more attention.

I was distracted by the Alaskan Bush People. Did you know you can't get pizza from under a log?

PairTheBoard

I think we will find that it doesn't take much to do the emotional stuff (the typical emotions like happiness, fear and anger along with the homeostatic emotions like hunger and pain) - the brain structures responsible for those sorts of thing are extremely small (and are included in lizards) and mostly hardwired.

Speech production is a bit more difficult - Broca's area isn't tiny and it takes a really long time for a human to learn how to communicate. Processing visual, auditory and touch inputs is exceedingly difficult and uses large amounts of brain matter (but we will most likely not design systems as inefficient as the brain for such tasks).

If we are going to be silly and pretend that capacity = number of connections between neurons in the brain (or number of connections x neuron firing rate), then I'd put it at far less than a thousandth.

Some interesting thoughts here. Thanks for the post.

An example of something that would require more power to become intelligent is the OpenWorm project. Currently they are attempting to simulate a nematode worm, which has 305 neurons in its brain. Obviously we don't think of a worm as intelligent at all, but if you can simulate that, then scale it up to simulating a brain with ~100 billion neurons you have a human equivalent.

You can't scale like that. Different parts of the brain are highly specialized.

I'm not suggesting multiply a worm brain by 300 million to get a human brain, I'm trying to show how computing power could lead to general intelligence when there doesn't appear to be any current computers that are even slightly intelligent. Obviously there is a lot of scientific research and software work necessary in addition to hardware, but first you have the capacity to simulate a worm brain, then a fruit fly, then a honeybee, then a mouse, then eventually you get to a human brain.

In case you missed it, the singularity is here:

http://www.dole.co.jp/wearablebanana/

Sexy! Who knew?

This is how I'm already being viewed by some. Singularity ready?

forumserver.twoplustwo.com/62/bbv4life/poast-pic-how-you-imagine-poaster-above-looks-laik-1488508/index5.html#post46151897

Son of the American Dream

(Obscure WWE reference)


PTB

You think this is intellectual enough for me? http://www.wwe.com/

Let's get one of those Hot Tub Time Machines for SMP.


PTB