Half man Half machine and AI

[kalesh]

I was reading an article about artificial intelligence. It gives some interesting forecasts

By 2030, a $1,000 computer system will have the power of 1,000 human brains; by 2050, 1 billion human brains

.

 

It also states using nanobots to expand human intelligence. Although I don't fancy placing nanobots in my brain. I would like to see the views of members of hypnography.

 

Here's a link to the article.

[Queso]

It's true.

 

Us monkeys have created a way to evolve intelligence.

 

We will use machines and AI to further evolve our own human intelligence, and spiral ourselves throughout the galaxy.

[pgrmdave]

I suppose it depends on how you rate the human brain. Currently, my computer can perform many, many, many more calculations than I can per second. However, it can't compose music. It can't learn new things. It can't apply data from one thing learned to another without being explicitly told to do so, and how to do it. It can't even make a decent cup of coffee. Is it really smarter than me just because it can do more math? Until computers can learn and be creative, they aren't anything more than (highly) glorified adding machines.

[God's servant]

See this is where you have to see just how smart the human brains are... i liked the way pgrmdave put it. And here's a thought... there are over 6 billion people in the world, over half of them are somewhat civilized, yet we still haven't found a way to get rid of world hunger, find a cure for cancer etcetera... so i ask you, "Is a computer with the "brain" power of 1 billion brains really that much better than what we have now?

[Queso]

Yeah it is. We will use the machines to evolve our own intelligence.

[God's servant]

But if its only as powerful as 1 billion brains compared to the 6 billion we have now then how will it help us to evolve. The computer only knows as much as what you program it to know. And then how do we know exactly how smart those brains are? Is it an average of how smart the population of the world is? So i ask you, "How can something that only knows as much as we program it to know help us evolve, we already know everything it knows?

[kalesh]

But if its only as powerful as 1 billion brains compared to the 6 billion we have now then how will it help us to evolve. The computer only knows as much as what you program it to know. And then how do we know exactly how smart those brains are? Is it an average of how smart the population of the world is?

 

We have 6 billion people but are any two people working on exactly the same problem at the same time? Usually not.

Now consider 1billion people working on solving a signle problem at the same time and on top of that they have telepathy(whatever one person thinks all of them know at once) The problem is sure to come to an end.

 

So i ask you, "How can something that only knows as much as we program it to know help us evolve, we already know everything it knows?

 

This is supposed to be an AI, in that we will program it to "learn" and it will then learn and program itself. If this takes place we will no longer be programming them. Instead it will be self conscious and learn through its experience. and through experiences of other computers as data is shared by next generation internet. At least I hope it happens.

[Queso]

Artificial intelligence will become far more advanced than we can imagine in the years to come.

It will take us beyond ourselves.

[Buffy]

But if its only as powerful as 1 billion brains compared to the 6 billion we have now then how will it help us to evolve.

Of course you have to read what dave said more carefully: a computer is not even as powerful as one brain in many respects. And of course evolution would take a while, and the machines themselves would likely evolve along with the changes of the brain to accomodate them.

 

And of course, this is the kind of thing that is incredibly hard to *direct* and is fraught with problems associated with unintended consequences and the moral quandries of the travails of the "failed branches" of this experiment in Eugenics.

The computer only knows as much as what you program it to know.

Not really, we do have "learning" programs. And of course defining what "learning" actually is is part of the problem.

 

Computers are currently pretty primitive, and the one thing that people really miss is that the brain is not a "computer with 100 billion bytes of memory" but rather a "massively parallel computer with 100 billion *processors*"

 

Von Neumann its not.

 

So,

So i ask you, "How can something that only knows as much as we program it to know help us evolve, we already know everything it knows?

That's because machines can indeed learn (start another thread if you want to discuss it), and within the scope of the idea postulated in this thread, it would be assisted by a human brain, thus becoming a tool to magnify its power just like a hammer or, whadda ya know, a computer!

 

Contact lenses for neurons,

Buffy

[Boerseun]

Interesting thread!

 

Wild speculation it might be, but there is an upper limit to calculation speed and storage capacity. Mainly, this is due to the 'grainyness' of matter, where you can only have as many individual elements in any matrix as there are atoms. In other words, if the individual 'cell' banking one bit of data reaches the atomic level, you're stuffed - no more advances in storage capacity is possible. All you can do is hope to develop a denser medium. The other limitation, of course, is the speed of light. Calculations in processors are already made faster than the data buses could transmit the results. With the end effect that the physical size of motherboards and the distance between devices is limited and must get smaller and closer together to make the processor work efficiently. It's useless to receive an anser in one millisecond but it takes two milliseconds to transmit over the bus, another two to verify the delivery, etc. Bring the device closer and it might only take one millisecond.

 

So, minaturisation of computer components isn't only due to increase portability, there's a definite need for it as dictated by the speed of light. And no, I'm not joking! It's true! But then you can only make the components smaller to a point where you're once again stopped by the innate 'graininess' at the cellular level.

 

I think the claim made in the original post is based on things like extrapolations of Moore's Law, which certainly cannot and will not hold. The Laws of Physics simply forbids it.

 

But that being said, computers will certainly get faster and smaller and more useful, but only up to a point. And whether they'll ever get to emulate or better the human brain is up to the future to decide. It's certainly possible, I guess. If you think about it, around 1,200cc's worth of gray matter encapsulated in the human skull should be possible to recreate on an electronic basis. The problem, in my mind, lies not so much with appliccable computer technology, but rather with our lacking understanding of the workings of the human mind.

[Buffy]

The problem, in my mind, lies not so much with appliccable computer technology, but rather with our lacking understanding of the workings of the human mind.

And there you have it! As I said, its really dangerous to think of the human brain as being "like" our current uniprocessor computers.

 

The point here is that the "processors" of the brain--the neurons--are incredibly slow and relatively (a term used advisedly, more below) simple. That just goes to show you what you can do with parallel computing! You got 100 Billion (with a "B") PROCESSORS up there! Wow! That's why its so incredible!

 

Now imagine enhancing that with a bunch of similarly simple (but massively faster and more capable) nanoprocessors running around "enhancing" those neurons. Now you're talking a *really* powerful "hammer for brains"!!!

 

Now, at the same time, we also have no idea how the brain actually works. I personally think we first have to figure this out and use this knowledge to completely change the computer architectures we use before we even think about how we might put these two things together. So unfortunately, at this point we're speculating about something where we can't even conceptualize the interactions, so the result of such speculations are, well, highly speculative!

 

Safety goggles required,

Buffy

[Jay-qu]

That is true boerseun, but what about quantum computers, or a computer somehow built on the manipulation of quarks - there are many possibilites that have yet to be entertained.

[kalesh]

I think the claim made in the original post is based on things like extrapolations of Moore's Law, which certainly cannot and will not hold. The Laws of Physics simply forbids it.

 

No it isn't. Please check the link in original post and read the article.

 

Now, at the same time, we also have no idea how the brain actually works. I personally think we first have to figure this out and use this knowledge to completely change the computer architectures we use before we even think about how we might put these two things together. So unfortunately, at this point we're speculating about something where we can't even conceptualize the interactions, so the result of such speculations are, well, highly speculative!

 

The article does mention using nanobots to map the brain...........

 

That is true boerseun, but what about quantum computers, or a computer somehow built on the manipulation of quarks - there are many possibilites that have yet to be entertained.

 

Light based quantum computer around the corner 1 Billion times faster than todays Supercomputers. Read more here

[Buffy]

Just a general commentary on the article: this thing is really old (dated 1999), and Ray's thoughts on this subject have been discussed to death. Ray's a smart guy, and I've spent a lot of money on his products over the years. While he's done some good work in the past on image recognition and machine translation, he's not really considered a leading light in computer science, and these speculations that he published are a major reason: most folks feel that at some point he just went off the deep end.

 

There's nothing really *wrong* with what he's talking about, other than his projections of "computer power" do indeed rely on Moore's law and make broad assumptions about the Von Neumann-ness of the human brain that are long out of the vogue, but he's gotten so far ahead of the technology--as I've discussed above--that its in the realm of pure science fiction: its cool to speculate about it, but its wild guesses that really have little basis in reality and there is no "probability" you could assign to "if" or "when" any of this would happen: saying we'll have nanobots in our brain in 2050 is not far from saying that we will have broken the speed of light by 2050: it *could* be true, but it sure looks unlikely right now, and there's no data or theories to even begin to back up such speculation.

 

Cool, but I wouldn't place any bets,

Buffy

[swampfox]

Personally, I think the neuron is qualitatively different than a transistorized processor. The neuron is non-linear, changes over time, has history, and are massively fed-back to one-another in the cortex. I'm saying this because of kalesh's original statement about computers having the power of human brains: human brains are qualitatively different than computer processors. It is this difference that has prevented AI from reaching it's full potential. A qualitative change will have to occur in technology, a new device developed, that more closely models a neuron and neural assemblies before we can begin to grow an artificial mind and start to compete with the human faculties we most admire: "observant, thoughtful, passionate, able to manipulate in the mind the symbols of language and mathematics, both the visions of art and geometry and poetry and music".

 

Edit: Oh yea, just though I'd mention the Blue Brain Project in case some don't know already and are interested. It's a simulation of the cortex which I believe will be based on the hodgkin-huxley model. I understand they plan to include history later.

[Queso]

I'm going to be so jelous when a robot can just spit beautiful poetry like it's nothing.

 

"IT COMES FROM THE SOUL"

 

Knot.

[pgrmdave]

The beauty of creativity is that it is not standardized - what is beautiful poetry to you may be crap for me and vice versa. A robot would no more be able to reliably compose beautiful poetry than a human - it would simply be able to do it quicker.