How long to human equivalent AI

The human brain is not organized anything like a modern computer.

The human memory is not organized like a computer database. It is more like an incredible number of hyperlinks. There is no index in the computer sense of the term.

A human brain can not be viewed as having software and hardware. It is like a computer which is constantly being rewired.

A human brain does not store data, it rewires itself. Furthermore, we know that the stored data relating to something like a Rsoe bush is scattered around the brain in a strange fashion. The outline of the green petals is stored with other outline dat, while the green & red colors are stored someplace else. The words relating to the rose bush and all its characteristics are simlarly scattered. The memory of how to say those words is stored separately.

If you read about the functioning of people with various types of brain injuries, you might be amazed. For example, some people can recognize colors visually, but cannot say the words red, blue, green, et cetera.

Considering the apparent speed of individual components, it is very fast at certain operations. A computer usually takes longer to decide that a searched for item is not there. A human can often say I never saw that guy/girl before almost immediately, while it might take a while to remember somebody you knew many years ago. This suggests some fundamental difference in organization.

We have very little knowledge of how to build a computer system which would function like a human brain, and there is eveidence that se cannot do it with raw processing power and huge memory resources.

The computers who beat chess experts can hardly do anthing else.

 

Log in or Sign up to hide all adverts.

RESTORED ?!......................

electroimage/visual transformation i,e.,etc.

 

Log in or Sign up to hide all adverts.

Yeah.
There is so much we don't know about how the brain works (well at least, there is so much, I don't know). It seems like most of what we do know is along the lines of: If we stick a fork in this part of the brain, he smells chocolate, but if we stick it in this part over here, he loses the ability to recognize human faces.

Hopefully we wont really have to understand it totally, in order to make it (like the approach of the research at Ai, that company in Israel, developing the HAL language system ).

Anyway, I'm always sceptical of the recent timelines saying 30 to 60 years and we'll have human level AI. We might have the technology to make it happen, but the knowledge might elude for longer than that. I always like to keep in mind that all futurists seem to say 30 - 60 years for this or that. But where's my flying car damn it!!!

 

Log in or Sign up to hide all adverts.

if we are first to assume that the human brain is fully functional outside of the body- say put a brain in a jar on feed it output, it would spit back responces like a computer, I would say that computers won't be able to reach human levels of intelligence untill each bit of computer memory is also a proccessor.

What I mean is that having one central proccessor and lots of memory like we do in today's machines it limited by the fact that you can only look at 32 or 64 bits of memory at a time. a given desktop pc today can only do one thing on at maximum 64 bits of data. that just a string of ones and zeros. imagine trying to learn some new programming technique if you could only keep one word in your mind at a time. you'd have to link each one to the other, and re-reference each word every time to read a new word, just to remember what you had just read. and in doing so, you would forget the new word!

until each one (on) and each zero (off) in the machine memory is represented by the proccessor states ( processing (on) and idle (off) ) of a billion small proccessors, we won't be even close to mimicing the human brain. Don't forget that while a neuron may fire @ about 200htz, it has access to many more nuerons, who are also all fireing at 200htz- and important note-not all in sync! the variation of neuron frequency from one area of thebrain to another may hold more information that we realize, and there is no physical item that we can disect to reverse engineer.

We're still long way off, IMO

 

We are not going to get human level AI using faster versions of current systems with more memory and other resources. Lots of parallel systems hooked to a huge common memory is a start, but that is probably not the way to go either.

Moores's law will fizzle out sooner or later. Nothing exponential can keep up for ever, not even for a long time.

What we need is an incredible new concept thought up by a genius. This could happen tomorrow or 200 years from now. Our current technology is not even close to any architecture that will do the job, and to repeat: More of what we now have is not the way to go.

Neural nets might be the answer, but that technology is in its infancy.

Remember that there have been a lot of failed technologies that held great promise. One outstanding example is cryogenic computers. Many bright people thought that systems operating near absolute zero with superconducting circuitry would be the wave of the future. Some small prototypes were actually built and tested, indicating that the concept had a lot of potential. Huge amounts of money were poured into that research, which has gotten nowhere and has been given up as a lost cause.

If I were told to develop human level AI, the best idea I could think of would be to spend a lot of money, time, and resources trying to determine what the brain of a 7-8 month old fetus is like. If you could simulate that electronically, it might learn for itself in less time than it takes a human brain to mature. Seems a lot easier that trying for an adult brain.

We might never develop human level AI due to practical considerations.

• Suppose it cost 5 million or more in current US dollars. A human is cheaper than that.
• Suppose you cannot get human level AI without some of the undesirable characteristics of a human being. I do not want my computer talking back to me and refusing to do boring accounting work.
• Suppose you have to let the system learn like a human, and cannot predict what its potential will be? AI is great if you can order ten engineers, 25 accountants, and some actuaries. What do you do with the artists, writers, and playboy personalities?
• Suppose, you have to contend with 2% being very clever electronic psychopaths that cannot be diagnosed until they actually do some serious damage. Does anybody remember an equipment breakdown that blacked out about 6-8 states for a day or so 20-30 years ago? If it had happened at night, there would have been aerial carnage from New York to Boston. During the day, all the pilots flew visually and left the crowded skies over New York almost as soon as they realized there was no air traffic control. A malevolent AI device could wreck havoc.
• Suppose, intelligence comes with the lower level of reliability common to most humans.

Intelligence is a tricky capability with a lot of potentially devastating unintended consequences.

I do not really believe in most of the above practical problems, but the cost effectiveness issue could kill human level AI. A human being is a damn capable device, and parents pay for most of their training and education (Id est: The programming comes cheap).

 

Maybe we can run it on Windows 95. That should ensure smooth transaction among digital neurons. However, you maybe forgetting no matter however powerful the computer it is still linear. It can't make decisions on it's own and god knows we don't have such knowledge yet to to make a analog program and i doubt we will in another 20 years. Programming has made great strides in the last 20 years but to programming a enviroment stable enough to chart the human brain is a astronomical task. We'll wait and see....i doubt we'll live long enough to see your proposed ideas come to life, we'll probably kill ourselves before that happens.


P.S. "will we still throw hissy fits even if we are in computers"

 

How long till human AI???? Never

As much as the technology develops it is far to much a task to even get the AI of a infant down let alone a human being. How do you account for creativity. If Computers do reach human equivilance in computing power can they write a "Romeo and Juliet", can they come up with a language of their own and utilize it in a efficient way. Can they paint, or even think of up a concept of a "the Sisitine Chapel". Getting to the point of having the same processing capabilities of a human brain is highly improbable (infact i think it's downright impossible) but even if it gets done it is still not the same as getting the AI to be human. It'll take unimagined skills in math and engennering to achieve what's proposed here and then keep in mind htat just achieveing normal, mediocre intelligence. How would you account for your Einsteins and Mozarts and Steven Hawkings. Will the Computer really be able to form his/her own consiousness and ponder about it's enviroment. I mean a scanning a picture of it's enviroment and putting it together to make it's own conclusions of what is there will itself require umparalled amounts of processing power but then it also has to que into the audio sensory around it too and combine it with the optical images ina less than a second.....see what i mean...getting it to do the simplest task will require major efforts. Face it guys, nature has us beat in the Ai department. It created a compact, 4lb, mother of all processing machines, and gave it to us for free. . So why not make regular humans the regular way. It's easy and ask anyone it's a lot of fun

Please Register or Log in to view the hidden image!

.

 

dinosaur
very enlightening posts
thanks

 

From what I have heard mentioned before it's what would be called a Meta image or a Layer.

The idea is that the brain works not with full images but patterns.

For instance if you think of someones face (someone that is close to you) you might immediately think of them with no problems, thats because at that time there are multiple layers that exist and the combined amount isn't "Fragmented".

(I say fragmented as I mean it in regards to the original Fragmenting of some content on a harddrive)

The Longer you are away from that person you thought of, the more you brain fragments with new influences. So the image starts to get less and less like them, and you will find yourself having to "Process" more to fill in the gaps and make the illusion of their face whole again.

The reason for this is when you process the image, it's like a Dot to dot in the brain, and you have to draw all the numbers together.

This is why the human brain can recognise multiple persons faces, and will build something similar to a compositing library from those that you know to save space.

This is also why it's very easy to mistake somebody for someone else.

 

I believe that AI is coming faster than we expect it to, because of the Technological Singularity. Obviously, we first need a machine that'll hold the equivalent of a human brain, and then perhaps we could copy a brain by activating all the possible reactions in the brain and putting these reactions into a simulation of somekind.
Also, once we create AI, what's to stop that intelligence from creating a higher intelligence, and so forth, and so on, until an intelligence beyond human, beyond even superhuman is created.
Hopefully there's still enough time for me to have some fun before society is warped by a super intelligent artificial entitiy.

Gurglo

 

Hey, hey Cris!

Glad to see you're still hanging in there.

At least some things haven't changed over here.

Take care.

:m:

Please Register or Log in to view the hidden image!

:m: