posted 06-14-99 11:13 AM ET         

We can only do so much, and we, human beings are the limit in the "equation." We can create randomness. But we'll never create a self-aware machine, even if teaches itself and makes decisions by itself, at the most basic level, there will always be a human involvment (computer code). My opinion: it is not and won't be possible.

posted 06-14-99 11:43 AM ET            

However, this intelligence will be like a bad clone of human intelligence (and never sentience, whatever that really means). If the U.S. semiconductor industry has its say, speed will substitute for lack of sophistication. There are a couple of limitations to speed, since it is difficult to etch smaller semiconductors. There are some novel etching processes out there, so I don't expect a grinding halt to Moore's Law, only a slowing. Also heat, etc. It is assumed that superconducting will hit the mainstream, so the effects of this barrier may be limited somewhat.

posted 06-14-99 09:57 PM ET            

There's nothing amazing about color perception based on R, G, and B components. Computers can do this now, and probably much better than humans. Perhaps you are trying to say something different, and I just didn't get your example?

threeover: "My opinion: it is not and won't be possible."

Human beings already create things that are more complicated than any single individual can understand. Even if the creation of a sentient machine was beyond the capabilities of a single person, why do you say it is impossible? What problems in particular are the stumbling blocks?

"However, this intelligence will be like a bad clone of human intelligence (and never sentience, whatever that really means)."

Why not sentience?

"If the U.S. semiconductor industry has its say, speed will substitute for lack of sophistication."

Brute force is my favorite algorithm.

Provost: "We are not governed by equations"

Sure we are. All physical laws can be expressed in terms of equations. Are you suggesting that human thought somehow violates the laws of physics?

JohnIII: "We are machines though. Just organic ones."

posted 06-15-99 01:14 PM ET            

"or can a deterministic computer program be considered sentient/intelligent?"
I don't know. Can it pass the Turing Test (cue Zekkei)

posted 06-15-99 01:40 PM ET         

Humans perceive the ability to communicate as being one of the major requirements for intelligence. A machine intelligence may not even consider humans to be �intelligent�, let alone want to communicate with them. A machine intelligence would, likely, operate on a vastly different time scale than humans. That would make human/machine communication, or any other interaction, very difficult.

posted 06-15-99 01:42 PM ET         

Once basic survival is taken care of, most human activity is devoted to filling emotional needs and wants. Many of those needs and wants derive from the fact that human intelligence �lives� in an organic body. So, what would a machine intelligence want or need? A machine intelligence would not have the instincts that go along with an organic body so it�s actions may not even be recognizable to humans as the actions of a �living� being. Fear is a powerful motivation to humans with the fear of personal death being among the strongest. A machine intelligence would probably not fear death.

Humans perceive the ability to communicate as being one of the major requirements for intelligence. A machine intelligence may not even consider humans to be �intelligent�, let alone want to communicate with them. A machine intelligence would, likely, operate on a vastly different time scale than humans. That would make human/machine communication, or any other interaction, very difficult.

posted 06-15-99 03:54 PM ET         

�For example, if you have AI which can create other AI, and then that can create more�

posted 06-16-99 12:34 AM ET            

If there is true AI, and they can build and repair themselves, what reason would there be for us to continue to exist? Humans won't be any use to self-repairing and self-replication Androids. Robots may go on, but Humans will be obsolete.

posted 06-16-99 05:18 AM ET            

I don't understand why people think having intelligent machines will be bad for us??

quote:

---

Having machine intellegence won't be good to us human.
If there is true AI, and they can build and repair themselves, what reason would there be for us to continue to exist? Humans won't be any use to self-repairing and self-replication Androids. Robots may go on, but Humans will be obsolete.

---

What reason would there be for us NOT to continue to exist?? What reason is there now for us to continue to exist? You say humans will become obsolete, which implies that right now we somehow aren't obsolete and serve some purpose. How so? How is that purpose (whatever it is) taken away by self replicating AIs?

Or are you simply afraid that AIs will somehow try to wipe us out? If these intelligent machines are truly intelligent then on what purpose will they try to eradicate the human race?

posted 06-16-99 10:45 AM ET         

That is interesting. The main point of the computer designing computer IIRC was that after a few �generations� we would have a computer that was so complex that no human could understand it. In that context, �computer� referred to both hardware and software. So that is basically what I was referring to. I will admit that I do not keep up much with AI progress though. It�s about all I can do to keep up with Microsoft�s �progress�.

Philip McCauley,

�Why couldn't the machine simply ask...�

Show me the simple code!

Jig,

posted 06-16-99 11:00 AM ET            

We already build computers that are more complex than any single human can understand.

posted 06-16-99 02:08 PM ET         

�We already build computers that are more complex than any single human can understand�

Two very good points:

�We�... That�s one point. Humans are still designing the computers. True there are many computer aids to the design, but there isn�t a computer that designs a computer without human intervention.

The second point is that although these computers are �more complex than any single human can understand�, they still are not intelligent AFAIK.

It seems to me that �intelligence� will be a function of software more than hardware. Hardware technology has far outpaced advances in software. The computer that I am typing this on is at least 1000 times faster than the first machine that I had over 20 years ago. All of the software that I had for the first four computers that I owned would fit in the RAM of this machine! Yet it�s not intelligent. Further, the software development tools that I use today are not 1000 times better than what I had 20 years ago. And, as Tolls pointed out, it is hard to define intelligence. And, as I said, humans might not recognize non-human intelligence anyway.

posted 06-17-99 12:23 AM ET            

The second point is that although these computers are �more complex than any single human can understand�, they still are not intelligent AFAIK."

Exactly. The computers we design aren't even close to intelligent like us, and we can't even understand them. Therefore, it is a bigger leap to believe that a machine could be built which had the capability to improve itself easily.

The genetic algorithm approach does seem like our best bet at the moment. The hard part is coming up with parameterized hardware/software that can achieve what we want. We would have to construct a framework withing which the intelligence could work, or we'd be doomed from the beginning. The interesting thing about this approach, however, is that it would be very non-linear. If we did succeed in finding such a framework, it would likely be non-obvious to make one that was, say, 10% better.

"It seems to me that �intelligence� will be a function of software more than hardware."

I agree with that. However, some people do have a problem with the notion of a completely deterministic system having the quality of "sentience". It also means that the computer you are using right now could be intelligent if it was running the right program. Interesting concept. Would it then be murder to turn your computer off?

"Further, the software development tools that I use today are not 1000 times better than what I had 20"

An interesting thing to note here is that the modern trend in software development takes it further and further away from the capabilities of an evolutionary program. When things are written in machine code, the space of "legal" programs is very large (compared to the space of possible combinations of bits), and there would tend to be a "fitness gradient" to try to find the optimal program. However, it is exceedingly unlikely that one could easily write a object oriented program that had "evolutionary" parameters, because the space of legal programs is very small compared to the space of programs that somebody could write (I have trouble with many object oriented languages, and I presumably know what I'm doing ). Is this a problem, or am I not being clever enough in the way I'm thinking about "constructing" the candidate programs?

Provost Harrison: "They are not sentient, but have different uses."

Are you saying they can't be sentient?

posted 06-17-99 01:28 AM ET            

well, we now have the option of fusing circuits and body cells, although we don't know what to do with it. Soon, we might be able to control a machine through this connection. What is stopping us then from getting rid of the human body, taking the brain, and hooking it up into the machine. For example build a robot dog, and put in an actual brain of a dog. does that classify as artificial inteligence? What if the brain is cloned, or otherwise 'produced'?

we now return you to your regularly schejuled forum topic...

posted 06-17-99 10:22 AM ET            

But that's not a good thing. This is analagous to saying "the appendix is a triumph of evolution, no human designer would ever include an appendix in a person".

"I believe the number of "cells" used was fewer than current designs would involve as well."

I'm exceedingly skeptical of that. This is the researcher's assertion, but I didn't quite follow his logic. Ring Oscillators are not that hard to design (although I'm trying to figure out if there would be any problems from the questionable nature of the analog and parasitic properties withing the FPGA), and a circuit to distinguish between 1 kHz and 10kHz should be very simple to construct, especially with a clock (which is a standard input to any digital circuit). The question of building a circuit that can distinguish between two arbitrary sounds is a different story, but I'm also not convinced that this would be very complex.

"In any case, I agree that using simple digital systems aren't the best way to proceed"

posted 06-18-99 11:29 AM ET         

Octopus: Was there a particular article or paper that you and Tolls were discussing ? Do you have a URL to it?

posted 06-18-99 09:42 PM ET            

"If humans use only 10% of the brain, what does the other 90% do?"

posted 06-19-99 12:35 PM ET            

Also, why shouldn't machines fear death? In fact, that's one thing that might keep them from the ever-so-feared "AI kills all humans" thing- we're integral to running those computers and providing input and designing new computers and keeping the power plants running, etc.

Hence the inherent need for the laws of robotics outlined by Asimov.

I think everyone should not just read the FAQ on the meaning of life, but click on all the links as well. There's one explaining why the Asimov Laws would never work, and would cause the computer to either behave irrationally (get mentally ill) or all the "proscribed thoughts" would break free over the programming.

http://singularity.posthuman.com/AI_design.temp.html#pre_prime

Read the other parts too- great info on Seed AI's and stuff!

posted 06-19-99 12:49 PM ET            

What I'm pointing out is that coercions are "above" the AI, and are supposed to have authority over the AI, so that when they break, they totally screw up the system - including all the other coercions. It is possible to design an elegant rational system that deals with rational conflicts. I'd really like to see an elegant rational system that deals with broken pieces of specialized override code. It would be fun to watch it fail. Imagine a corporation where all the layers of management demand absolute obedience from those below, and refuse to listen to any questions. Now imagine giving all the managers LSD. It doesn't matter how smart the engineers are, that company is going to die. (In fact, who needs LSD? The company would probably die in any case from accumulated natural stupidity. Happens all the time.) And if the coercions aren't absolute, if they're just suggestions, then what's the point?

[snip]

The Principle of Irrationality: If you design an irrational system, it will become irrational in a way you don't expect.

posted 06-21-99 08:41 AM ET         

Saw �Matrix� over the weekend. We DO NOT need machine intelligence!

posted 06-25-99 08:51 AM ET         

Did our previous threads ever come to a conclusion as to what intelligence is anyway ?

As I see it, if you are of the opinion that human intelligence solely arises from the interaction of the neurons within your brain, then you must surely believe in the eventual creation of machine intelligence (if we, or some other species live long enough to perform the creation). It is just a case of achieving the ability to produce the level of complexity found in the human brain (or near to it). A massive neural network perhaps. This probably implies even machine intelligence has to evolve.

If, on the other hand, you think that the brain just acts to a) keep the maintenance tasks going for the body, and b) interpret the control of something more spiritual/soul-like then the whole situation becomes much more complex.

Initially it would seem as if you were doomed to achieve machine intelligence as the best you could achieve would be these maintenance tasks and responses that may look intelligent, but that has not intelligence behind them. But, if a soul/spirit is able to inhabit a carbon based body who is to say that it could not inhabit a silicon based one instead ?

Silicon heaven, after all where else would all the old calculators go to ?

So which murderer is going to pull the mains plug out ?

posted 06-25-99 01:56 PM ET         

Eh ? Oh you're referring to the 'Eternal Golden Braid' book are you. So much to read already, so little time...

posted 06-28-99 12:52 PM ET         

Meanwhile, "Wot no more posts ?"

posted 06-28-99 02:04 PM ET            

Does it mean the same as artificial flavors? An artificial flavor is a flavor which mimics another flavor. So, are we talking about mimicing human abilities? That seems awfully limiting -- why would you want something that is just the same as what you already have (people)?

It seems to me that what people think about when they think about AI is creations which have the abilities of computers (the ability to do repetitive tasks with a high degree of precision without tiring) with the abilities of humans (inductive reasoning, artistic flare, intuitive recognition of complex patterns like voices and faces, etc.).

posted 06-29-99 04:57 AM ET         

The definition isn't restricted to mimicking humans, but as this is arguably the best example of intelligence we have to date, it is a good place to start. We agree humans are intelligent (to a greater or lesser degree) so let's go for that.