Topic Actions

Topic Search

Who is online

Users browsing this forum: No registered users and 33 guests

Artificial Intelligence

Join us in talking discussing all things Honor, including (but not limited to) tactics, favorite characters, and book discussions.
Re: Artificial Intelligence
Post by DDHvi   » Thu Oct 08, 2015 8:57 pm

DDHvi
Captain (Junior Grade)

Posts: 365
Joined: Mon Dec 15, 2014 8:16 pm

I "understand" that the basic problem at present is that we don't understand understanding.

Sorry, I just couldn't resist that. But no one, at present, does know enough about understanding to use it in any machine. What we can do is use the understanding of the programmers to produce imitations or models of what real understanding would produce.

This might be the difference between a real Turing machine, and something that appears to be a Turing machine, as long as you don't go outside of its parameters.
Douglas Hvistendahl
Retired technical nerd
ddhviste@drtel.net

Dumb mistakes are very irritating.
Smart mistakes go on forever
Unless you test your assumptions!
Top
Re: Artificial Intelligence
Post by Tenshinai   » Fri Oct 09, 2015 1:18 pm

Tenshinai
Admiral

Posts: 2893
Joined: Tue Nov 02, 2010 8:34 pm
Location: Sweden

DDHvi wrote:I "understand" that the basic problem at present is that we don't understand understanding.

Sorry, I just couldn't resist that. But no one, at present, does know enough about understanding to use it in any machine. What we can do is use the understanding of the programmers to produce imitations or models of what real understanding would produce.

This might be the difference between a real Turing machine, and something that appears to be a Turing machine, as long as you don't go outside of its parameters.


While your initial statement does sound a bit silly, yes pretty much.
Top
Re: Artificial Intelligence
Post by bjchip   » Sat Oct 10, 2015 2:33 am

bjchip
Midshipman

Posts: 3
Joined: Sat Dec 27, 2014 10:17 pm

Tenshinai wrote:
DDHvi wrote:I "understand" that the basic problem at present is that we don't understand understanding....
This might be the difference between a real Turing machine, and something that appears to be a Turing machine, as long as you don't go outside of its parameters.


While your initial statement does sound a bit silly, yes pretty much.


We've been going about it wrong, is all. No AI that does not have a body that it and only it controls, a neural-net that learns to control said body, and an environment in which to use that body can ever develop "understanding".

Yet it can come in through the back door of a full virtual reality system if that is implemented with sufficient fidelity. The reason is that the *I* that differentiates us from our computer toys, and they are all toys under our control, is a product of the environment we don't control and the body we do control.

I'm working through this on-and-off and have been thinking about it for most of 40 years. At the moment "time" seems to be my difficulty but I will try to finish writing the thesis.

Several things come out of my version of this. One is an explanation for both sleep and dreams. Another is that an AI without a body will of necessity be insane in our terms, and finally that most mammals have an *I* concept... though most have nothing like our model of the world to go with it.

However, what I would commend to your attention is that if we can manage to build an AI (and I think we can if we do it correctly) we can also find a way to transfer human memories and understandings into a machine. The lines would become blurred. Eventually we will evolve to be something else... or die. I like the idea of surviving through evolution. :-)

Anyhow... I'm just an occasional visitor to this place. I love the books. I just worry about how much is still to be written. I'm sure David is as well.
Top
Re: Artificial Intelligence
Post by Tenshinai   » Sun Oct 11, 2015 11:52 am

Tenshinai
Admiral

Posts: 2893
Joined: Tue Nov 02, 2010 8:34 pm
Location: Sweden

bjchip wrote:We've been going about it wrong, is all. No AI that does not have a body that it and only it controls, a neural-net that learns to control said body, and an environment in which to use that body can ever develop "understanding".

Yet it can come in through the back door of a full virtual reality system if that is implemented with sufficient fidelity. The reason is that the *I* that differentiates us from our computer toys, and they are all toys under our control, is a product of the environment we don't control and the body we do control.


An interesting idea.

Personally, and looking at how attempts have gone, especially like how the once so hyped neural nets would solve eeeverything, i rather doubt that it will "fix" it.

But it´s still a very interesting idea to explore.

bjchip wrote:I'm working through this on-and-off and have been thinking about it for most of 40 years. At the moment "time" seems to be my difficulty but I will try to finish writing the thesis.


Please do. And drop a note here when you do.

bjchip wrote:and finally that most mammals have an *I* concept... though most have nothing like our model of the world to go with it.


Research in the last few decades have found a lot of animals to be much closer to humans than previously thought, we just need better frames of reference to get any sort of understanding across.

And you can find what appears to be quite well developed sense of self among some squids for example, so not just mammals, it´s just easier to make sense of there.

Another interesting part of that research though is that there may be a very large gap between individuals, someone did a study with common housecats some years ago and found that the variance in level of sentience and intelligence was extreme, with a rare few that was found was capable of near or possibly humanlevel understanding, while on the opposite extreme there was just a very noted absence of anything appearing intelligent or sentient.

It was also found that the most "aware" cats were almost always those who were treated more like children than like pets.
(which is something i have seen myself, my aunt has always had cats, and no matter how they are when they come to her, they soon develop very clear personalities and starts responding more or less intelligently over time)

bjchip wrote:However, what I would commend to your attention is that if we can manage to build an AI (and I think we can if we do it correctly) we can also find a way to transfer human memories and understandings into a machine. The lines would become blurred. Eventually we will evolve to be something else... or die. I like the idea of surviving through evolution.


Ghost in the shell... (which will get a live action movie added to the series in 2017 by the way)
Top
Re: Artificial Intelligence
Post by JohnRoth   » Sun Oct 11, 2015 4:00 pm

JohnRoth
Admiral

Posts: 2438
Joined: Sat Jun 25, 2011 6:54 am
Location: Centreville, VA, USA

DDHvi wrote:I "understand" that the basic problem at present is that we don't understand understanding.

Sorry, I just couldn't resist that. But no one, at present, does know enough about understanding to use it in any machine. What we can do is use the understanding of the programmers to produce imitations or models of what real understanding would produce.

This might be the difference between a real Turing machine, and something that appears to be a Turing machine, as long as you don't go outside of its parameters.


Amusing, and pretty accurate. One of the biggest problems, IMO, is that we're making it too complicated. Another is that linguistics is still headed off in the wrong direction at warp speed. A lot of linguists still think that the syntax of sentences is important.

The biggest breakthrough in linguistics in the last half century seems to be not very well known. It's not in syntax: it's in semantics. There are 65 "semantic primes," that is, basic concepts that everyone over the age of five knows intuitively and so don't have to be defined. Any so-called AI that doesn't have them built in as the base for all other conceptual structures won't "understand" in the same way humans do. Ref: https://www.griffith.edu.au/humanities- ... e-homepage
Top
Re: Artificial Intelligence
Post by Tenshinai   » Tue Oct 13, 2015 3:17 pm

Tenshinai
Admiral

Posts: 2893
Joined: Tue Nov 02, 2010 8:34 pm
Location: Sweden

A lot of linguists still think that the syntax of sentences is important.


Oh boy, did you really just say that syntax is irrelevant?

Seriously, try saying that again after you have experience with some languages where syntax is blatantly exact or where word position matters greatly.


Hyperbasic demonstration why:
Där är du, (oh) there you are

Där du är, where you are

Är du där, are you there?

Är där du, (terrible language that noone would use(someone might use Är det där du though)) is that you there?

Du är där, you are there (*points on map*)

Du där är, (incomplete sentence) you there is *something*


So, just changing word order with just the 3 same words, i still got 6 different meanings out of them. Even if being superstrict, there´s 4 100% valid meanings and one kinda.

Syntax is not something to leave out if you want to be understood.


The biggest breakthrough in linguistics in the last half century seems to be not very well known. It's not in syntax: it's in semantics. There are 65 "semantic primes," that is, basic concepts that everyone over the age of five knows intuitively and so don't have to be defined


Ehm, you do know that that idea has been long since flushed out of credibility? No, not everyone knows them and they certainly don´t know them "intuitively" either.

We learn them by exposure. With lack of interaction, or interaction with nonstandard surroundings, those "primes" may be partially or completely nonexistant in anyone.
Top
Re: Artificial Intelligence
Post by Loren Pechtel   » Tue Oct 13, 2015 8:33 pm

Loren Pechtel
Rear Admiral

Posts: 1324
Joined: Sat Jul 11, 2015 8:24 pm

And to those of us lacking a Christian bent this idea sounds horrifying.

cthia wrote:And I totally agree that your optimistic path is certainly possible. It is why several posts ago I stated that it might be quite wise to impart the idea of a Deity within its programming. Somehow the idea of an AI grounded in faith and Christian values, regardless of one's own beliefs, is less intimidating.

A Confuscious AI sounds welcoming, even...logical.
Top
Re: Artificial Intelligence
Post by JohnRoth   » Tue Oct 13, 2015 9:55 pm

JohnRoth
Admiral

Posts: 2438
Joined: Sat Jun 25, 2011 6:54 am
Location: Centreville, VA, USA

JohnRoth wrote:A lot of linguists still think that the syntax of sentences is important.


Tenshinai wrote:Oh boy, did you really just say that syntax is irrelevant?

Seriously, try saying that again after you have experience with some languages where syntax is blatantly exact or where word position matters greatly.


No, I didn't say that. What I said is that the fetish of thinking that the syntax of the sentence is the be-all and end-all of linguistic analysis is misguided. Noam Chomsky comes to mind as one of the primary exponents of that notion, and, if you'll notice, his theory has gone through at least four complete revisions and has never worked. It may be significant that Chomsky only speaks one language.

*irrelevant example deleted.

JohnRoth wrote:The biggest breakthrough in linguistics in the last half century seems to be not very well known. It's not in syntax: it's in semantics. There are 65 "semantic primes," that is, basic concepts that everyone over the age of five knows intuitively and so don't have to be defined


Tenshinai wrote:Ehm, you do know that that idea has been long since flushed out of credibility? No, not everyone knows them and they certainly don´t know them "intuitively" either.

We learn them by exposure. With lack of interaction, or interaction with nonstandard surroundings, those "primes" may be partially or completely nonexistent in anyone.



Please cite evidence that it has "long since been flushed out of credibility." Who said that, when did they say it, where can I find it other than in an academic library in languages which I don't know? The work I cite is fairly well regarded by modern semanticists, not that many of them don't have issues with one or another aspect.

As far as intellectual respectability is concerned, Anna Wierzbicka was awarded the 2010 International Dobrushin Prize by an institute of the Russian Academy of Sciences. This is no small honor. The English translation of her acceptance speech is here: https://www.griffith.edu.au/__data/asse ... People.pdf (Original in Russian.)

There is, by the way, no suggestion that interaction isn't important, but if that interaction exists, it will come in the same way that language comes. The one exception that's suspected to date is Pyraha, which seems to lack number words; it also seems to lack Chomsky-style recursion, which is interesting on a different level. The first hasn't, however, been studied by a linguist who is competent in NSM.
Top
Re: Artificial Intelligence
Post by Tenshinai   » Wed Oct 14, 2015 8:36 pm

Tenshinai
Admiral

Posts: 2893
Joined: Tue Nov 02, 2010 8:34 pm
Location: Sweden

No, I didn't say that. What I said is that the fetish of thinking that the syntax of the sentence is the be-all and end-all of linguistic analysis is misguided.


Nope, not what you said, you need to adjust your syntax. :mrgreen:

Seriously though, of course it´s not the only important part.

Please cite evidence that it has "long since been flushed out of credibility." Who said that, when did they say it, where can I find it other than in an academic library in languages which I don't know?


It´s not a matter of who and when, but rather by how many and how often.

You might not find it by looking for linguistical research however, your best bet is probably to look for research done on progression of child development.

The idea of the "original language" is extremely old, but have essentially been debunked as variation between cultures, ages and individual differences have shown big holes in it.

As far as intellectual respectability is concerned, Anna Wierzbicka was awarded the 2010 International Dobrushin Prize by an institute of the Russian Academy of Sciences. This is no small honor.


Never said she wasn´t good. But it´s her pet peeve and since it is essentially a "soft science", disproving it 100% is unrealistic, even though by now it´s probably at least at ~90% disproven.
A point where i just cannot accept it as likely any longer, no matter if it still has a fair number of supporters.

Basically, unless every single contraindicating research result so far ends up being proven wrong for some reason, the idea is dead in the water and sinking.

That´s not to say there´s absolutely no hints of truth in it, it has been shown that infants have a few concepts clear extremely quickly, but the relation to a "first language" is shaky.

Most likely, some of those concepts are there at birth, while others are gained by exposure.

And the basic idea that there are commonalites across languages, i agree completely with.

However, "first/original/natural/prime language" or the universal metalanguage like is described?

No. As long as there are ANY naturally occuring languages that completely lack one or several of those "universal" building blocks, that alone kills the idea stone dead instantly.


Looking at their list, concepts i know isn´t fully in *some* language:
Maybe
Because
Small OR Big(they use a scale with just either instead)
True (problematic to say the least)
One/Two (probably including Piraha)
Other

Also, pretty much all the time-related concepts have languages where they are more or less nonexistant.

And last time i read an article on anything relevant to this was probably last year so i currently have absolutely zero data to provide.

There is, by the way, no suggestion that interaction isn't important, but if that interaction exists, it will come in the same way that language comes. The one exception that's suspected to date is Pyraha, which seems to lack number words; it also seems to lack Chomsky-style recursion, which is interesting on a different level. The first hasn't, however, been studied by a linguist who is competent in NSM.


I assume you mean Piraha? I had to look it up before i realised what you meant.

And yes that´s one of the more interesting oddballs currently known. Not the only one though, but probably the most extreme.

I wouldn´t be surprised if they´re missing something when looking at Piraha though, the inconsistencies are too varied and off to just be random, "bad science" or fake.

There´s clearly a system there, and the people have used it successfully for quite a time, so my first guess would be to start looking for secondary information providers, like, are they also doing signs with their hands/fingers while talking? Are they using head movement as a modifier? Etc...
Top
Re: Artificial Intelligence
Post by JohnRoth   » Wed Oct 14, 2015 9:16 pm

JohnRoth
Admiral

Posts: 2438
Joined: Sat Jun 25, 2011 6:54 am
Location: Centreville, VA, USA

Tenshinai wrote:
JohnRoth wrote:No, I didn't say that. What I said is that the fetish of thinking that the syntax of the sentence is the be-all and end-all of linguistic analysis is misguided.


Nope, not what you said, you need to adjust your syntax. :mrgreen:

Seriously though, of course it´s not the only important part.

JohnRoth wrote:Please cite evidence that it has "long since been flushed out of credibility." Who said that, when did they say it, where can I find it other than in an academic library in languages which I don't know?


It´s not a matter of who and when, but rather by how many and how often.

You might not find it by looking for linguistical research however, your best bet is probably to look for research done on progression of child development.

The idea of the "original language" is extremely old, but have essentially been debunked as variation between cultures, ages and individual differences have shown big holes in it.


I'm finally getting an idea of what you think has been debunked. For the record, I have never thought of NSM as having anything to do with an "original language," if that term has any meaning, which I seriously doubt.

Tenshinai wrote:
JohnRoth wrote:As far as intellectual respectability is concerned, Anna Wierzbicka was awarded the 2010 International Dobrushin Prize by an institute of the Russian Academy of Sciences. This is no small honor.


Tenshinai wrote:Never said she wasn´t good. But it´s her pet peeve and since it is essentially a "soft science", disproving it 100% is unrealistic, even though by now it´s probably at least at ~90% disproven.
A point where i just cannot accept it as likely any longer, no matter if it still has a fair number of supporters.

Basically, unless every single contraindicating research result so far ends up being proven wrong for some reason, the idea is dead in the water and sinking.

That´s not to say there´s absolutely no hints of truth in it, it has been shown that infants have a few concepts clear extremely quickly, but the relation to a "first language" is shaky.

Most likely, some of those concepts are there at birth, while others are gained by exposure.

And the basic idea that there are commonalites across languages, i agree completely with.

However, "first/original/natural/prime language" or the universal metalanguage like is described?

No. As long as there are ANY naturally occuring languages that completely lack one or several of those "universal" building blocks, that alone kills the idea stone dead instantly.


Looking at their list, concepts i know isn´t fully in *some* language:
Maybe
Because
Small OR Big(they use a scale with just either instead)
True (problematic to say the least)
One/Two (probably including Piraha)
Other

Also, pretty much all the time-related concepts have languages where they are more or less nonexistant.

And last time i read an article on anything relevant to this was probably last year so i currently have absolutely zero data to provide.


I tend to go with adequately researched data rather than offhand comments.

As far as early childhood learning is concerned, it's one of the few areas that I give Chomsky positive credit for kickstarting. Again, I've never said that these concepts are present at birth - the data to date is not only that they aren't, they occur in sequence in the first few years. There are, in fact, some people who don't seem to get one of the last ones, "word" because they were brought up by deaf parents who never taught them sign language.

There's an apparently interesting book about such a person: "A Man Without Words," by Susan Schaller. I haven't read it, but it's referenced by Ian Tattersall, one of the leading researchers on human origins, in "Masters of the Planet: the Search for our Human Origins" (2012).

What's interesting about the early childhood learning material is that most children go through a stereotyped sequence of acquisition as long as there is a decent environment for them to learn from. They all learn the same things, in about the same order. If this isn't a classic definition of instinct, I don't know what is.

JohnRoth wrote:There is, by the way, no suggestion that interaction isn't important, but if that interaction exists, it will come in the same way that language comes. The one exception that's suspected to date is Pyraha, which seems to lack number words; it also seems to lack Chomsky-style recursion, which is interesting on a different level. The first hasn't, however, been studied by a linguist who is competent in NSM.


I assume you mean Piraha? I had to look it up before i realised what you meant.

And yes that´s one of the more interesting oddballs currently known. Not the only one though, but probably the most extreme.

I wouldn´t be surprised if they´re missing something when looking at Piraha though, the inconsistencies are too varied and off to just be random, "bad science" or fake.

There´s clearly a system there, and the people have used it successfully for quite a time, so my first guess would be to start looking for secondary information providers, like, are they also doing signs with their hands/fingers while talking? Are they using head movement as a modifier? Etc...


The data on Piraha looks weird to me, too, but it does seem pretty well established.

By the way, I'd appreciate it if you'd do me the courtesy of leaving my name in the quote blocks rather than removing it.
Top

Return to Honorverse