AI thread

Discussions on everything related to the software, electronic, and mechanical components of information systems and instruments.

Re: Artificial Intelligence thread

Postby wolfhnd on March 17th, 2016, 4:57 pm 

Inchworm » Thu Mar 17, 2016 7:24 pm wrote:No need for a move to be faster in the head than in reality. If things around us would move at the speed of sound for instance, no need to think at the speed of light. At first, the brain was made to move wrt actual stimuli. It still does, but it has developed a way to imagine future ones and to act as if they were actual. The parallel way the brain works treats more information per second than a computer though, so a fast treatment is not only dependent on the speed of the information. Moreover, information can take multiple directions at a time in a three dimensional array, whereas it can take only one at a time in one wire.


Computer use multi processors and brute force to make up for their lack of "eloquence".
User avatar
wolfhnd
Resident Member
 
Posts: 4631
Joined: 21 Jun 2005
Blog: View Blog (3)


Re: Artificial Intelligence thread

Postby Natural ChemE on March 17th, 2016, 5:31 pm 

wolfhnd » March 17th, 2016, 3:57 pm wrote:Computer use multi processors and brute force to make up for their lack of "eloquence".

Animals like humans do the same, except we call the processors neurons.

Wikipedia's got a list of the number of neurons in animal brains. The top three, in descending order:
  1. African elephants have 267 billion neurons.
  2. Humans have 86 billion neurons.
  3. Rhesus macaques (monkeys) have 6 billion neurons.
It's almost too easy to make a powerful AI with sufficiently many neurons; our main technological limitation is artificially producing enough neurons to be competitive with biological brains.

In July 2015, researchers reported an artificial neural network (ANN; basically an AI's brain) with 160 billion parameters - but that it had only three multi-core computers, so less than 1,000 neural processors.

To make up the difference in raw computational power, it's likely that the first strong AI's will be more eloquent than comparable biological brains. This should be possible due to computational diversity; while human brains are constrained in how diverse their computational nodes can be, AI brains can combine all sorts of computational units.

For example, how many telephone numbers can you memorize? Us humans are limited at these sorts of tasks because our brains are relatively homogeneous in their design; we lack massive data storage regions where we can memorize arbitrary data like telephone numbers. By contrast, AI can have massive data storage units for such simple memorization tasks. Ditto for stuff like mathematical calculations, physical intuition, etc.

Our strong AI will have neural networks that think like we do. They'll just also have all sorts of other thinking structures in addition to those that we have.
Natural ChemE
Forum Moderator
 
Posts: 2754
Joined: 28 Dec 2009


Re: Artificial Intelligence thread

Postby wolfhnd on March 17th, 2016, 5:45 pm 

Natural ChemE » Thu Mar 17, 2016 9:31 pm wrote:
wolfhnd » March 17th, 2016, 3:57 pm wrote:Computer use multi processors and brute force to make up for their lack of "eloquence".

Animals like humans do the same, except we call the processors neurons.

Wikipedia's got a list of the number of neurons in animal brains. The top three, in descending order:
  1. African elephants have 267 billion neurons.
  2. Humans have 86 billion neurons.
  3. Rhesus macaques (monkeys) have 6 billion neurons.
It's almost too easy to make a powerful AI with sufficiently many neurons; our main technological limitation is artificially producing enough neurons to be competitive with biological brains.

In July 2015, researchers reported an artificial neural network (ANN; basically an AI's brain) with 160 billion parameters - but that it had only three multi-core computers, so less than 1,000 neural processors.

The first strong AI will come about only after technology becomes sufficient to provide the computational power necessary to make them. Since it's a competitive field, it's likely that the first AI brains will be more sophisticated than human brains but with less raw power.


Fruit flies have 250,000 neurons in their entire nervous system. Eloquence in design is the key to "common sense" the kind of intelligence needed to avoid getting eaten and to find a mate and successfully copulate.

Other processes can be accomplished by brute force but we are talking at the moment about the kind of autonomy a fruit fly has not abstract problem solving. It seems that a mix of the two is most desirable as autonomy is perhaps important to independent learning in some way. I don't know if that kind of autonomy is necessarily related to the kind of robots that manage a household. Some people think it is some don't. There are ways to make a pretty stupid robot "safe".
User avatar
wolfhnd
Resident Member
 
Posts: 4631
Joined: 21 Jun 2005
Blog: View Blog (3)


Re: Artificial Intelligence thread

Postby hyksos on March 29th, 2016, 5:14 am 

Inchworm » March 17th, 2016, 11:24 pm wrote:. The parallel way the brain works treats more information per second than a computer though, so a fast treatment is not only dependent on the speed of the information. Moreover, information can take multiple directions at a time in a three dimensional array, whereas it can take only one at a time in one wire.


I posted this article in its own thread here at the forum. Nearly a month later it has received no replies. Maybe it will get more airplay if I move it into this thread.

http://sentic.net/jumping-nlp-curves.pdf

As you have pointed out, information is moving through the 3dimensional bulk of the brain, and that information is repeated in its branching. There is a particular passage from the above PDF regarding this issue.

As human text processors, we do not have such limitations as every word we see activates a cascade of semantically related concepts, relevant episodes, and sensory experiences, all of which enable the completion of complex NLP tasks—such as word-sense disambiguation, textual entailment, and semantic role labeling—in a quick and effortless way.

(Emphasis added by me. )

What is most interesting about this article is the following. There is still a contingency of Ai researchers, mostly young ones, who believe that meaning can still be found by having a statistical algorithm mine statistics out of a large corpus of text. I have always been very strongly opposed to this. The meaning of the word "tree" is not ever contained in how often the word "tree" appears in proximity to other words in english text. The fascinating thing about this article is that the author not only says why this does not work : but then he goes onto explain exactly why. The brain is a highly connected network. A word will quote, "activate a cascade" of neuronal activity that is roughly speaking the semantically related concepts.

In any case, there is definitely some important issue going on here, regarding the difference between computer circuits and mammalian brains. The issue here seems to be the level or amount of connectedness between units. The whole phrase "activate a cascade" really has so many sideways implications and intriguing metaphors. Unfortunately, the author backs away from this analogy almost immediately, and goes back to speaking in a more business-like manner.
User avatar
hyksos
Active Member
 
Posts: 1321
Joined: 28 Nov 2014


Re: Artificial Intelligence thread

Postby Natural ChemE on March 29th, 2016, 1:01 pm 

hyksos,

You've brought up an excellent point about text-mining AI. Timely, too, since Microsoft's new chatbot Tay was taken offline just 16 hours after launch when it seemed to become a neo-Nazi. Of course Tay wasn't a real neo-Nazi because it didn't even know who Hitler was; it just connected words like "Hitler" to phrases based on user input - a behavior that trolls quickly exploited.

We refer to text-mining algorithms as a type of AI, but they're fundamentally different from how strong AI and human intelligence work.

PS - Ya know, thinking about it, the term "AI" is so broad that it might be prudent to come up with new terms for general use. In practice this is already possible by qualifying AI as a "strong AI" or "semantic AI" or whatnot, though I suspect that these names suggest more similarity than actually exists.
Natural ChemE
Forum Moderator
 
Posts: 2754
Joined: 28 Dec 2009


Re: Artificial Intelligence thread

Postby Braininvat on March 29th, 2016, 2:01 pm 

I think finding certain kinds of implied meaning are going to be a huge challenge for AI. The other day, the example of ironic usage came up, and I wondered how an AI would discern the way that many people now use the word "fascinating." In many contexts, the word is now used in sarcasm. So there's a whole complex minefield of social clues that come with a word usage like that. On message boards, when someone conveys the sarcastic usage, they may rely entirely on context and the listener's understanding of who they are as a person....or they may put "fascinating" in a different font, italicize, use 20 pt. letters, etc.
User avatar
Braininvat
Forum Administrator
 
Posts: 6589
Joined: 21 Jan 2014
Location: Black Hills


Re: Artificial Intelligence thread

Postby neuro on April 1st, 2016, 10:09 am 

hyksos » March 2nd, 2016, 8:55 pm wrote:I had hoped to communicate that trial-and-error approaches lack immediate simple inferences that First-order logic agents perform naturally and easily.


I'm not so sure that First-order logic agents are so clever and versatile. What seems to be so "immediate simple" generally is a specific instance (or logic rule) that has been injected into the system. Deriving the logic rule itself is not equally simple...

hyksos » March 3rd, 2016, 5:03 am wrote:This all very natural for us to handle mentally, because we live, think and talk in this manner. No research project in Ai has any really thorough-going methodology to capture this type of reasoning. One marginal mention here would be SOAR agent (invented by John Laird).


I'm once more not so sure that "this manner" is anything different than having a thousand circuits examining the situation in 1000 different ways (kind of trial and error approach) and, by a mechanism of selective attention based on relevance and success, picking the best solution. The only difference, in our brain, is that this is done simultaneously - 1000 different ways of elaborating the same info - because we have enough neurons and synapses to do it. This, however, has already been said by Natural ChemE
User avatar
neuro
Forum Moderator
 
Posts: 2635
Joined: 25 Jun 2010
Location: italy


AI, its erratic history and possible futures.

Postby Event Horizon on May 14th, 2018, 11:19 am 

An interesting article here on its history and it's trajectory. It seems that it is now having a renaissance and is a major part of a mainly military arms race. This article gives us a potted history, potential futures and lots of interesting links. You'll need to allocate a fair bit of time to this article if AI is your thing, but it's very interesting.

https://warontherocks.com/2018/05/its-e ... -learning/
User avatar
Event Horizon
Member
 
Posts: 302
Joined: 05 Mar 2018
Location: England somewhere.


Re: AI thread

Postby Braininvat on May 14th, 2018, 11:44 am 

Merged with existing AI thread. Enjoy.
User avatar
Braininvat
Forum Administrator
 
Posts: 6589
Joined: 21 Jan 2014
Location: Black Hills


Re: AI thread

Postby zetreque on May 14th, 2018, 12:43 pm 

I have come to the conclusion that all this AI stuff is dumb as hell.
Humans already have a hard enough time accepting other religions, cultures, ethnicities, sexes, and species. Humans fear aliens or any force more powerful than themselves. Why on Earth are there now thousands of people interested in contributing to develop a far superior form of life to themselves? One that can compute far faster, the result is unpredictable, and will make humans mostly obsolete. Human laziness might be their own undoing.

PS: another tesla crash was in the news. Last I heard they don't know if autopilot was in use or not.
User avatar
zetreque
Forum Moderator
 
Posts: 3642
Joined: 30 Dec 2007
Location: Paradise being lost to humanity
Blog: View Blog (3)


Re: AI thread

Postby zetreque on May 14th, 2018, 12:54 pm 

Call me crazy if you want but I am meeting so many people these days that are wanting to participate in AI and with all that I am seeing I think there is actually a very real possibility we could see some version of the terminator movies in my life time. Minus the liquid metal guy and time travel maybe.

I will officially call out anyone working in this field that wants to do more than just improve the operation of a remote controlled vehicle or some other basic used of AI an idiot from now on.
User avatar
zetreque
Forum Moderator
 
Posts: 3642
Joined: 30 Dec 2007
Location: Paradise being lost to humanity
Blog: View Blog (3)


Re: AI thread

Postby hyksos on May 15th, 2018, 7:20 am 

Why on Earth are there now thousands of people interested in contributing to develop a far superior form of life to themselves? One that can compute far faster, the result is unpredictable, and will make humans mostly obsolete.


King Midas didn't get what he wanted. He got exactly what he asked for.


User avatar
hyksos
Active Member
 
Posts: 1321
Joined: 28 Nov 2014


Re: AI thread

Postby zetreque on May 15th, 2018, 2:52 pm 

Even if you can teach ethics to a car or other machine, how can you force one set of ethics across society over another set of ethics? Isn't that itself unethical? Isn't that comparable to forcing your religion on others through dangling a fancy technology in front of their face? More than that, you aren't just dangling a fancy technology in front of them, you are forming a society where it's hard to function without that technology and therefore they have to accept it without that much of a choice. It also seems like another step toward forming a monoculture and less diversity.
User avatar
zetreque
Forum Moderator
 
Posts: 3642
Joined: 30 Dec 2007
Location: Paradise being lost to humanity
Blog: View Blog (3)


Previous

Return to Computers

Who is online

Users browsing this forum: No registered users and 8 guests