What on the Earth would AI do?

  • 70 Replies
  • 23978 Views
*

Don Patrick

  • Trusty Member
  • ********
  • Replicant
  • *
  • 633
    • AI / robot merchandise
Re: What on the Earth would AI do?
« Reply #15 on: July 06, 2014, 01:59:42 pm »
That would be the grand question of Natural Language Processing, an entire scientific field, heavily intertwined with the equally big question of how to represent knowledge, and no less a challenge than making a universal translator.
I don't have an answer, but I should point out that some parts of language don't follow any detectable pattern, like rare Old English spellings.
CO2 retains heat. More CO2 in the air = hotter climate.

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #16 on: July 06, 2014, 07:53:45 pm »
Just brainstorming a little bit here...

So, we have a unit description mechanism. But how to pair its structure with certain input, as input should build an unit representation? We would have two distinct representations, one for description language (input) and one for imagined units in the Universe (knowledge from input).

User writes input, the algorithm builds a unit representation and finds out how to improve current situation. I guess that basic Logic should be tightly integrated to the mechanism. Like in sentence :

If Ann is less than 6 years old she should go to kindergarten, but if she is more than 6, she should go to school.

This unit representation would be:

Child (
    Name {@String} ["Ann"],
    Age {@Int},
    GoingTo {
        Kindergarten |
        School
    }
    [
        @Age < 6 => @KinderGarten;
        @Age >= 6 => @School
    ]
)

So we would have if-then, and, or and not hard-coded in instinct and learn somehow their representation in language we are using to communicate with machine. Algorithm could guess what certain words mean, check it out by further observation or by asking questions and according to all past guesses build a unit representation that can be further analyzed for improvements.

So far, so good, knowledge is based on statistical analyzing of input, it starts from knowing instinct and pairs instinct notions with certain input particles and builds units in knowledge.

But, input particles are tricky to recognize from zero knowledge about them. At least we know that all different languages have similar particles like subject, object and predicate. So again, we can pair subject-object-predicate with certain input particles based on statistical analysis and questioning teacher.

At the end we would have
1. knowledge about languages (subject-object-predicate)
2. knowledge about language->unit representation (if-then, and, or, not)

I don't like that subject-object-predicate distinction. We can represent all of them by predicates with zero, one, two or more parameters. Subject would be a predicate with zero parameters.

And what if we have just basic notions of our unit storing mechanism in instinct (sets, variables, sequences, choices, booleans and reactive values - see Synth specification overview (http://aidreams.co.uk/forum/index.php?topic=6549.0)). Then we should pair these notions with whatever input we get. I like this idea more than previous ones. The problem is now reduced to statistical analyzing of input stream and checking conclusions either by observing, either by asking right questions. We have to check every combination of different input interpretation and remember those that gave positive results in recognizing.

I think some progress is made with this post, thank you guys :)

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #17 on: July 07, 2014, 02:20:21 pm »
Basic recognition would be like this: if we have two inputs

1. Mother cooks lunch
2. Father cooks dinner

Algorithm should  conclude following rule:

set Sentence (
    Subject {Mother | Father},
    Predicate {Cooks},
    Object {Lunch | Dinner}
)
[
    [Subject [Mother], Predicate [Cooks], Object [Lunch]];
    [Subject [Father], Predicate [Cooks], Object [Dinner]]
]

Word cooks repeats, we investigate what is near and get sentence pattern. I think that with enough input we could automatically conclude whole English grammar (up in round braces) and we would get specific knowledge about the world (down in square brackets). With each new inputted sentence we would have to revise our grammar. So that would be sequences and choices solved (see Synth Specification Overview).

Now we have to find out how to recognize other forms of knowledge (sets, events and reactive values). After recognition mystery solved, we will have a system that recognizes units from input. Then we have to define how to automatically improve recognized units - that is what would machine do, it would improve the world.

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #18 on: July 09, 2014, 12:11:45 pm »
Behavior mechanism could be implemented as a switching mechanism between these 3 parts:

1. recognizer
2. learner
3. planner and doer

Recognizer at 1D + time sense such is text input could be programmed exactly the same as a top-down text parsing algorithm (we start from the most general notion and go down, level by level to more specific notions). When an input is not recognized, learner should be invoked to produce new, yet unknown rules. When the input is finally parsed we can move on.

Behavior mechanism should control when to do which of those three parts (recognize, learn or plan and do things). Behavior mech. should be constantly refreshed with new plans upon recognizing and learning things from the world. All three parts could be done with automatic, "unconscious" process (instinct), while switching between those three parts would be implemented as a "conscious" process (decisions according to learned and recognized facts).

Right now I'm having trouble with learning new rules. It should be done with genetic algorithm that blindly composes new rules and then checks do that rules apply to the input. A lot of rules would be imagined, but only few of them would actually hold the truth. It is easy to add new rules to the parser, we just put a new "choice" or a new "tuple" and the input parses. The problem is that majority of new composed rules would be rubbish.

So, how to eliminate rules that are fake? Maybe some statistical analyzing, but I'm not sure how exactly it would implemented.

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #19 on: July 09, 2014, 06:23:12 pm »
... There are no "fake" rules. If they stand either for one example then they are good, but not very usable. It is like induction - the more amount of examplars they cover, the better quality they are. So each rule should have some statistical data about amount of examplars the rule parsed in the past. Those with higher degree of use are more likely to hold and should be checked first when recognizing things. Also, when creating new rules (learning), genetic algorithm should combine first those with higher degree of use...

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #20 on: July 15, 2014, 02:57:43 pm »
... but let's suppose that we have a statistical unsupervised learning system that translates textual input into internal representation of units inside Universe. We would have a lot of little systems in format:

set System (
    TimeStamp {@Date},
    Input {@String},
    set Representation {@Synth}
)
[
    [-Time1-, 'daughter cooks breakfast', @SPO["daughter", "cooks", "breakfast"]],
    [-Time2-, "if it is rain, it is wet but if it is sunny, it is dry", @SPO [it is rain |= it is wet; it is sunny |= it is dry],
    ...
]

But how to decide what to do now? What guides would machine have, what to do next? This system can hold knowledge and answer questions, but intelligent behavior is another thing. Maybe answer would be optimization of the Universe, but how it would be performed? What are units purposes and strivings?

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #21 on: July 16, 2014, 11:58:44 am »
... each unit can influence other units and that is its purpose or striving. I.e. other unit has if-then pair that concludes from the first unit. Now we can detect conflict (when two different unit states are required in the same time). In this case some alternative solution should be proposed. Or we can eliminate or replace units in a sense of optimization of the Universe. Again, proposition should be brought up...

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #22 on: July 22, 2014, 01:15:34 pm »
...about learning...

The machine could learn a language by guessing what word means what, then by statistically checking does that word fulfill its imagined role in other places. This way any natural language could be learned from the scratch, but what is with other means of communication? What if an AI is infiltrated in weird company of creatures that only solve math tasks and nothing else?

Animals do not use language, yet they function fair well. I wonder how they think...

*

Freddy

  • Administrator
  • **********************
  • Colossus
  • *
  • 6856
  • Mostly Harmless
Re: What on the Earth would AI do?
« Reply #23 on: July 22, 2014, 02:49:47 pm »
Animals do communicate with sounds - couldn't that be language ?

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #24 on: July 22, 2014, 03:04:42 pm »
so basically it should be the same thing as with humans...

*

Freddy

  • Administrator
  • **********************
  • Colossus
  • *
  • 6856
  • Mostly Harmless
Re: What on the Earth would AI do?
« Reply #25 on: July 22, 2014, 03:38:55 pm »
If that's the case and an AI could learn human language basically just by observation; then could it learn what the animals are talking about ?  :o

*

Don Patrick

  • Trusty Member
  • ********
  • Replicant
  • *
  • 633
    • AI / robot merchandise
Re: What on the Earth would AI do?
« Reply #26 on: July 22, 2014, 08:46:36 pm »
Many animals do use language, here is one professor who learned the language of prairy dogs through observation: http://www.treehugger.com/natural-sciences/researcher-decodes-praire-dog-language-discovers-theyve-been-calling-people-fat.html
Dolphins also use a language we have yet to decode completely, though researchers have learned that dolpins call eachother by name.
Ants communicate through pheromones, though these are more a form of instructions.
Birds, no doubt, talk all the time, they don't make all that noise just for fun.
Dolphin language is one of the greatest secrets of the universe, and particularly hard to analyse, so it could well do to put a "big data" AI to the task of finding consistent patterns.

As for robots, I know of several projects: http://www.popsci.com/technology/article/2013-02/watch-childlike-humanoid-robot-begin-comprehend-language

However, the successful robot projects are based on learning through hands-on training, not by putting a man/AI in China without a Rosetta stone and expecting him/it to come out writing perfect Chinese.
CO2 retains heat. More CO2 in the air = hotter climate.

*

Art

  • At the end of the game, the King and Pawn go into the same box.
  • Trusty Member
  • **********************
  • Colossus
  • *
  • 5865
Re: What on the Earth would AI do?
« Reply #27 on: July 22, 2014, 09:42:39 pm »
Good one Don!!  :2funny:
In the world of AI, it's the thought that counts!

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: What on the Earth would AI do?
« Reply #28 on: October 08, 2014, 05:51:07 pm »
Well, Chinese babies are born into China, yet some of them manage to write perfect Chinese books later in their lives, right?

So I believe an AI could do the same.

*

Don Patrick

  • Trusty Member
  • ********
  • Replicant
  • *
  • 633
    • AI / robot merchandise
Re: What on the Earth would AI do?
« Reply #29 on: October 08, 2014, 07:19:23 pm »
It could, if it had teachers.
CO2 retains heat. More CO2 in the air = hotter climate.

 


OpenAI Speech-to-Speech Reasoning Demo
by MikeB (AI News )
March 31, 2024, 01:00:53 pm
Say good-bye to GPUs...
by MikeB (AI News )
March 23, 2024, 09:23:52 am
Google Bard report
by ivan.moony (AI News )
February 14, 2024, 04:42:23 pm
Elon Musk's xAI Grok Chatbot
by MikeB (AI News )
December 11, 2023, 06:26:33 am
Nvidia Hype
by 8pla.net (AI News )
December 06, 2023, 10:04:52 pm
How will the OpenAI CEO being Fired affect ChatGPT?
by 8pla.net (AI News )
December 06, 2023, 09:54:25 pm
Independent AI sovereignties
by WriterOfMinds (AI News )
November 08, 2023, 04:51:21 am
LLaMA2 Meta's chatbot released
by 8pla.net (AI News )
October 18, 2023, 11:41:21 pm

Users Online

213 Guests, 0 Users

Most Online Today: 359. Most Online Ever: 2369 (November 21, 2020, 04:08:13 pm)

Articles