Author
|
Topic: Machine Evolution (Read 7119 times) |
|
Cassidy McGurk
Adept
Gender: 
Posts: 128
Reputation: 7.62
Rate Cassidy McGurk
http://www.isec.info/ get me out of here!
|
 |
Machine Evolution
« on: 2002-08-07 03:50:15 » |
 |
Regarding the discussion with the Friendly AI promoters I was reminded of my favourite quote from George Dysons Darwin Among The Machines
"In the game of life and evolution, there are three players at the table: human beings, nature, and machine. I am firmly on the side of nature. But nature, I suspect, is on the side of machines."
|
I must remember to change this sig regularly
|
|
|
Ophis
Magister
Posts: 176
Reputation: 6.26
Rate Ophis
|
|
Re:Machine Evolution
« Reply #1 on: 2003-01-12 11:49:41 » |
|
We had an interesting chat about the question of machine, evolution, and artificial intelligence on the BBS recently. I had to sign-off at some point and the topic hadn't been settled to my liking.
The first question was wether we thought that AIs would one day surpass human intelligence. On this topic, everyone seemed to have been in agreement that yes, the day would come when an AI would be able to learn and make intellectual judgements better than human beings.
The second question was about how distant that future was. Some proposed 50 years or so, whereas others were of the opinion that it would take much longer, perhaps 150, or more, years.
I was personally of the second opinion. When debating this, the question was spinned differently: What is it that is missing in today's AI? Are we simply missing sheer processing power to be able to match human intelligence? Or is there something more fundamental that's been left out or that has yet to be discovered?
On this question, I was also of the second opinion. I think that as good as AI engines may be at playing chess and at pattern recognition, I think that we are missing something fundamental in today's AI.
The adaptability of the human brain suggests to me that we might be missing a notion or concept on how we define the rules that govern a system's ability to process information.
Maybe what we're missing is just this: how do we bring subjectivity to an AI? Subjectivity would appear to be a critical element in the development of self-preservation, opinions, creativity, and other instincts and even emotions that are certainly cornerstones to biological intelligence as we know it.
Are instincts simply rules coded in our brains or genes? Does the same go for emotions? What makes a 10 months old infant giggle and laugh at the sight of Lucifer's cute little Brussel Griffon dog? I'm afraid that, on these topics, I have more questions than answers... Does anyone have any insights? Books to recommend?
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
|
|
Re:Machine Evolution
« Reply #2 on: 2003-01-12 23:08:18 » |
|
Quote from: Ophis on 2003-01-12 11:49:41
The second question was about how distant that future was. Some proposed 50 years or so, whereas others were of the opinion that it would take much longer, perhaps 150, or more, years.
|
That is not my recollection and LogLady agrees. I compared AI to landing on the moon and asked if others thought we were closer to Goddard launching model rockets or da Vinci drawing pictures of flying machines. You and sat are the only ones that answered, and you both said we are closer to da Vinci. I didn't mention it at the time but I think we are closer to Goddard.
Later we were talking about the acceleration of technological progress. I commented, "I guess we saw more progress in the last 100 years than the previous 1000", then asked, "I'm interesting in how everyone would fill in X: We will see more progress in the next X years than in the entire 20th century." You answered "10".
Let's say you are correct and we are as far away from AI as da Vinci was from landing on the moon, about 500 years of progress. How long it will take us to make the same amount of progress in the future? If, as you said, we will see more progress in the next 10 years than we saw in the entire 20th century, it would be pretty safe to say that we will see AI in less than 20 years.
|
|
|
|
|
Ophis
Magister
Posts: 176
Reputation: 6.26
Rate Ophis
|
|
Re:Machine Evolution
« Reply #3 on: 2003-01-13 00:43:25 » |
|
As wonderful as the human brain may be, mine often tends to play tricks on me when it comes to information archival and retrieval.  Appologies for the misrepresentation of facts.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
|
|
Re:Machine Evolution
« Reply #4 on: 2003-01-13 11:11:34 » |
|
Quote from: Ophis on 2003-01-13 00:43:25
As wonderful as the human brain may be, mine often tends to play tricks on me when it comes to information archival and retrieval. |
Mine too, that's why we need external mind tools like the chat log 
So do you agree that your two answers ("da Vinci" and "10") lead to the conclusion that we will see real AI in less than 20 years?
|
|
|
|
|
Ophis
Magister
Posts: 176
Reputation: 6.26
Rate Ophis
|
|
Re:Machine Evolution
« Reply #5 on: 2003-01-14 00:52:50 » |
|
Call me a pessimist but I'm not seeing any sort of quality general-purpose AI for another 100-150 years. But then again, I don't know much about the field so that's why I was asking those questions above.
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
|
|
Re:Machine Evolution
« Reply #6 on: 2003-01-14 01:54:45 » |
|
Quote from: Ophis on 2003-01-14 00:52:50
Call me a pessimist but I'm not seeing any sort of quality general-purpose AI for another 100-150 years. But then again, I don't know much about the field so that's why I was asking those questions above.
|
Maybe Ray Kurzweil's Law of Accelerating Returns will turn you into a realist (which probably looks like an optimist if you're a pessimist).
|
|
|
|
|
David Lucifer
Archon
Posts: 2642
Reputation: 8.94
Rate David Lucifer
Enlighten me.
|
|
Re:Machine Evolution
« Reply #7 on: 2003-01-14 02:33:10 » |
|
The computational power of the human brain has been estimated at 20 PetaFLOPS (based on 100 billion neurons * 1000 connections/neuron * 200 calculations/connection/s).
I was curious about how the world's fastest supercomputer compares, so I went to Top500 site where I found this interesting graph. The top line is the sum of the performance of the top 500 supercomputers. The middle line is the fastest supercomputer and the bottom line corresponds to the last computer on the list.
It looks to me like if the current trend of increasing power 1000 fold/decade continues, we will see a single system with the equivalent computational power of the human brain in about 10 years.
|
|
|
|
|
Ophis
Magister
Posts: 176
Reputation: 6.26
Rate Ophis
|
|
Re:Machine Evolution
« Reply #8 on: 2003-01-15 17:33:22 » |
|
Interesting... I'll make sure to add a few AI books to my reading list. I might yet be convinced :-)
|
|
|
|
|
Hermit
Archon
Posts: 4287
Reputation: 8.94
Rate Hermit
Prime example of a practically perfect person
|
|
Re:Machine Evolution
« Reply #9 on: 2003-01-20 17:46:50 » |
|
Quote from: David Lucifer on 2003-01-14 02:33:10
The computational power of the human brain has been estimated at 20 PetaFLOPS (based on 100 billion neurons * 1000 connections/neuron * 200 calculations/connection/s).
I was curious about how the world's fastest supercomputer compares, so I went to Top500 site where I found this interesting graph. The top line is the sum of the performance of the top 500 supercomputers. The middle line is the fastest supercomputer and the bottom line corresponds to the last computer on the list.
It looks to me like if the current trend of increasing power 1000 fold/decade continues, we will see a single system with the equivalent computational power of the human brain in about 10 years.
|
I would tend to agree with Lucifer - with some reservations and caveats. After all, given what is leaking from DARPA (refer http://www.darpa.mil/) these days, systems coming quite close to this capability already exist - and if that can leak, it does not take a vast stretch of the imagination to suspect that such a system or systems may already exist but remain cocooned in red tape.
I couple similar reasoning and understanding to Lucifer's with Moore's law as previously modified (the original was too pessimistic) and the fact that knowledge of the needed techniques is so widespread that the instantiation of such systems has already become inevitable as processing capacities increase and the tools improve (also practically inevitable if our society's production capability survives long enough). So barring our sudden, massive destruction, I'd suggest that single gifted/genius level human equivalence will be available in a single $5,000 desktop computer (2003 US$) within 20 years (or sooner). And the equivalent of the processing capability of all humans within 50 (or as seems likely to me, probably very much sooner if the delivery and availability of the single genius equivalence results in a further acceleration of computer capability/capacity) - also on a $5,000 desktop system.
Would this make us humans into slavemasters - or redundant (or is that redundant too?) - unless we change a lot in the same period. Not impossible. But extremely unlikely without social/political upset that will make the fall of Rome to Christianity look positively mundane. Assuming we make the transition, whatever will humans do with their time (especially given that responding to Deesian floods of stygian proportions could then be totally automated... <grin>)? The machines would be so much better at finding creative solutions than us that we would not really be needed - except possibly as audiences. But a simple minded bot could quite likely replace us there too - and much more efficiently.
Kind Regards
Hermit
|
With or without religion, you would have good people doing good things and evil people doing evil things. But for good people to do evil things, that takes religion. - Steven Weinberg, 1999
|
|
|
Ophis
Magister
Posts: 176
Reputation: 6.26
Rate Ophis
|
|
Re:Machine Evolution
« Reply #10 on: 2003-01-21 00:05:54 » |
|
I was reading the "Basic Questions" section of the document that is referenced from the Artificial Intelligence wiki page (http://virus.lucifer.com/wiki/ArtificialIntelligence) wherein John McCarthy answers two questions that express some of my doubts about the creation of a human-level AI system in the near (20-50 years) future.
Mr. McCarthy doesn't mention any timeline in his answers but he does mention that many researchers think that there is something missing in today's AI research.
I guess I'm just a little more pessimistic than you guys in my estimations of when we'll discover these missing "new fundamental ideas".
Quoted from http://www-formal.stanford.edu/jmc/whatisai/node1.html:
Q. How far is AI from reaching human-level intelligence? When will it happen?
A. A few people think that human-level intelligence can be achieved by writing large numbers of programs of the kind people are now writing and assembling vast knowledge bases of facts in the languages now used for expressing knowledge.
However, most AI researchers believe that new fundamental ideas are required, and therefore it cannot be predicted when human level intelligence will be achieved.
Q. Are computers fast enough to be intelligent?
A. Some people think much faster computers are required as well as new ideas. My own opinion is that the computers of 30 years ago were fast enough if only we knew how to program them. Of course, quite apart from the ambitions of AI researchers, computers will keep getting faster.
|
|
|
|
|
Hermit
Archon
Posts: 4287
Reputation: 8.94
Rate Hermit
Prime example of a practically perfect person
|
|
Re:Machine Evolution
« Reply #11 on: 2003-01-21 02:31:48 » |
|
[Ophis] I guess I'm just a little more pessimistic than you guys in my estimations of when we'll discover these missing "new fundamental ideas".
Quote:Q. How far is AI from reaching human-level intelligence? When will it happen?
A. A few people think that human-level intelligence can be achieved by writing large numbers of programs of the kind people are now writing and assembling vast knowledge bases of facts in the languages now used for expressing knowledge. |
[Ophis] However, most AI researchers believe that new fundamental ideas are required, and therefore it cannot be predicted when human level intelligence will be achieved.
Quote:Q. Are computers fast enough to be intelligent?
A. Some people think much faster computers are required as well as new ideas. My own opinion is that the computers of 30 years ago were fast enough if only we knew how to program them. Of course, quite apart from the ambitions of AI researchers, computers will keep getting faster. |
[Hermit]
I would suggest visiting http://www.imagination-engines.com/thaler.htm, and noting in particular both the "Clientele" which includes amongst others (including a number of commercial entities):- The US Air Force Research Laboratory, Materials and Manufacturing Directorate, Wright-Patterson Air Force Base
- Munitions Directorate, Eglin Air Force Base
- NIST
- Advanced Refractory Technologies, Buffalo, NY
I find the applications list (partially cited) highly suggestive of the capabilities of the systems which have been realized to date:- autonomous materials discovery
- the invention of products and services (i.e., personal hygiene products)
- neural networks that write their own computer code
- control systems for chemical vapor deposition reactors
- autonomous classification
- self-prototyping devices
- artificial life
Before or after that, you may find reviewing the following threads helpful
You might like to abbreviate the above by focusing on the first page of the first link, and my responses (as I generally include, precis or indicate the text posted by others in my replies.)
|
With or without religion, you would have good people doing good things and evil people doing evil things. But for good people to do evil things, that takes religion. - Steven Weinberg, 1999
|
|
|
|
|
|
|
|
|
|
Logged
signo-pregunta.jpg
