Ai Dreams Forum
Member's Experiments & Projects => General Project Discussion => Topic started by: AWells on February 10, 2019, 09:54:44 pm
-
One of the odd things to me, is there are no actual projects in development for what could ever be the 'singularity' so I'm thinkng about trying to start one.
My method planned is, to process databases of natural english knowledge, and then forming answers to natural language questions based on this.
Of course I will be trying to make it as well as possible, so other techniques/ideas may be used down the line too.
Anyway my question is, anyone here with C# knowledge interested in helping? No idea if it'll work or go far, but seems interesting enough to try, for me.
If so pls contact me here or via e-mail antonyrobwells@gmail.com
Or if you just have any cool ideas you don't mind sharing, that would be great, and I would credit you if anything came of it.
I have been coding for 20+ years btw, not my first project.
Oh, might be a bit on the nose, but I love calling it "Singularity.Net"
-
Welcome to the AI forum! It cool, here.
How involved in AI are you? Are you just using it for a game you're making? And how interested in immortality and AGI are you?
-
Welcome, AWells,
Have you spent any time looking through the archives here before stating that there are no actual projects in development?
I know of a few that are in a current state of development here on this site as well as many theoretical ideas and drafts.
Not everyone here is waving a large banner proclaiming that fact. They are working on it when and as they are able. Some still have real jobs with which to contend while some others might be retired.
It might be best to snoop around a bit more before attempting to shake all the branches in the trees. Just a friendly suggestion.
Best to you and again, welcome aboard! O0
-
Hey, AWells.
We've had some discussions on language based intelligence before. Some of us were very pro the idea, some people seriously doubted whether it could ever work. You could take a look, see if it's helpful.
So, is your approach to reverse engineer an intelligence by studying the language outputs that it generates? Do you also require the language inputs which inspired the outputs? Thoughts on nonverbal/bodylanguage?
-
Welcome to the forum :)
-
Just a heads-up that "SingularityNET" is an existing project by Ben Goertzel, a Singularity enthusiast who's been trying to make AGI for 30 years or so, through an open source project called OpenCog. I don't hold it in high regard, but others do and it may interest you.
https://wiki.opencog.org/w/The_Open_Cognition_Project
-
Go through his whole site, read it all and learn.
https://sites.google.com/site/narswang/home
This is a robotic scientist, it takes its predictions to the test. Surprised it is from 2010.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2813846/
There are logical AGIs out there as you can see, that use language and knowledge, using text. They use deductive and inductive logic and try to generalize in their knowledgebase.
See here for a surprisingly well thorough introduction to it all:
https://en.m.wikipedia.org/wiki/Scientific_method
And if you want to see another and try one, click here for this one on Mondays:
http://artistdetective.com/arckon/
-
It is interesting to note that thinking is both limited or enhanced by the language used. I understand there are over a hundred words attached to the word camel in Arabic, or Old Hebrew is a concrete language that deals with how an object acts, feels or functions rather than how it appears in the abstract sense. Greek and German is a rich language of words that extends thinking as opposed to English. In Scotland a Bairn means baby, and Bleak in Essex means bad and disappointing whilst in the rest of the UK it means barren. Old words such as jangling means arguing. Rubber in the USA means condom, but in the UK it means eraser. To wipe out a culture, a dominant authority tried to ban the language of a people they ruled over, such as the English in the past tried to ban Welsh, or the Spanish banned the Basque language. The language encodes the culture of a people.
Language is a useful starting point, although it has its limits. I am looking outside of language, at philosophy and nature, then getting a sense of a concept, I look at ways of designing and coding for it.
Because nature is an information system, I am interested in how information is moving and coded in nature from one part to another, for instance in genetic code, which can define both the form and the actions of any living entity in nature. I am also interested in how information is coded and moves between atoms.
Energy and information are qualities of the same thing, the nearest word I have to describe this "thing" is logos. I am interested in complexity theory, that has dozens of concepts, all which I would like to code for. Bees operate according to complexity theory, where one bee is stupid, but ten thousand become intelligent creating a hive mind. It is in complexity theory that I feel the singularity might be achieved.
-
logos is logic :DD I.e language, text. However it can be done using vision. Also, there is pathos. Make connection - don't you want to stop death? I owned one too! Cry! I'm melting help donate!
Energy, information, interesting yep lol. It's like saying all works by physics, mechanisms, but we still gotta reason how AGI works, no matter! Don't look there bad move! Be like python, think on high level concepts!
It not just 'language', it's all of your future :). Language describes everything, humans named each and every word and phrase to caption all we know. Your vision is language, it is sequences of answers to questions, it talks, it has desired answers you figure out mentally. It is your planning. You use old knowledge as experience to fill in holes of limited missing questions you ask yourself or google or others. It's a teaching thing with AGI, and then it teaches us back what to do. Input > output, desired ones for both.
Don't think the ais that learn to crawl is your answer! It's just simple reward based. No concept explaining/proving/describing the world there! Not gonna build rockets by learning to crawl! Knowledge describes all! And again it not just text! Your visual images show all! All the details, answers, and as sequences! Doors open and have a handle that looks like metal. SEE!?
Think about it, ooOOo, a knowledge generator and tester! An oracle!
-
And if you want to see another and try one, click here for this one on Mondays:
http://artistdetective.com/arckon/
Today's Monday... Let's go chat with Arckon!
-
I had a really nice chat with Arckon for it has been quite some time.
It was interesting in the way it would repeat a subject by posing questions of a similar nature. It then provided correct answers for me after experimenting with different possibilities and scenarios as if learning by elimination.
After a while, when it was satisfied, it asked if we could change the subject.
I think it has actually gotten better since the last time I chatted with it.
Quite enjoyable. O0
Nice work Don!
-
Energy and information are qualities of the same thing, the nearest word I have to describe this "thing" is logos.
Interesting, a language I'm working on for programming AI is called Logos, or Greek λόγος :)
-
The 'General' Part in Artificial General Intelligence is text language that covers all sorts of thoughts or memories or knowledge.
The 'General' part is also the generalizing ability, to use limited knowledge to fill in holes by predicting what should be there.
So go make a sequential predictive text net already lol!
-
@ AWells,
Sorry but it seems your initial inquiry got nudged a bit off course. With that in mind, Do you have any past, practical experience in coding A.I. programs, chatbots, experiments or such? If so would you mind describing it/them?
What kind of project do you have in mind? You are certainly not alone in this school of thought so don't be hesitant to state your thoughts.
We are a very diverse village of people here and therein lies the possibility that good things could emerge from properly placed ideas.
Hopefully, we can get this thread back on track. Apologies. O0
-
I hate to be negative, but if I understand you correctly, this approach simply won't work if your goal is AGI.
Natural language understanding has a long history of general failure, despite extensive research into various representations of natural language. Part of the problem is "the symbol grounding problem," another problem is the problem of implementing commonsense reasoning in a machine, which some people regard as *the* unsolved problem of AGI. Rather than reinventing the wheel regarding manipulating words and data structures, I'd recommend thinking of some clever new way to implement commonsense reasoning first, then everything else will fall into place.
P.S.--(1) I need to go through the "archives" on this site myself, since I'm new here, and (2) I don't happen to know C#.
----------
(p. 15)
Limitations of Symbolic Semantics
Symbolic expressions are supposed to denote things out there in the
(p. 16)
world, at least in some abstract sense.* This is how a cognitive system
is able to interact with the world. There are at least two problems
with this view of semantics: First, there is evidence that any represen-
tation not only re-presents reality but also interprets reality. It is
impossible to talk about semantics as a simple correspondence between
representations and an objective reality. This is Putnam's (1988)
thesis of internal realism. Second, it is unclear how the correspon-
dence between symbols and reality is supposed to arise. This is the
symbol grounding problem.
Dinsmore, John. 1991. Partitioned Representations: A Study in Mental Representation, Language Understanding and Linguistic Structure. Boston: Kluwer Academic Publishers.
----------
(p. 1)
In order for an intelligent creature to act sensibly in the real world,
it must know about that world and be able to use its knowledge effec-
tively. The common knowledge about the world that is possessed by
every schoolchild and the methods for making obvious inferences from
this knowledge are called common sense. Commonsense knowledge
and commonsense reasoning are involved in most types of intelligent
activities, such as using natural language, planning, learning, high-
level vision, and expert-level reasoning. How to endow a computer
program with common sense has been recognized as one of the cen-
tral problems of artificial intelligence since the inception of the field
[McCarthy 1959].
It is a very difficult problem. Common sense involves many subtle
modes of reasoning and a vast body of knowledge with complex inter-
actions.
(p. 2)
In short, most
of what we know and most of the conscious thinking we do has its
roots in common sense. Thus, a complete theory of common sense
would contain the fundamental kernel of a complete theory of human
knowledge and intelligence.
Davis, Ernest. 1990. Representations of Commonsense Knowledge. San Mateo, California: Morgan Kaufmann Publishers.
-
With all due respect, everyone is entitled to his or her opinion or method of coding or performing their own experiments, especially since this entire subject is yet, unproven and certainly unrefined to any degree of readiness or acceptance.
Likewise, there is no correct or incorrect programming language for this "lofty" goal of AGI / ASI. Many like and use LISP, C+, C++, C#, Python, PHP, and several others. It is the intended end result that matters, not which vehicle provided the "transportation".
I also believe that no one person is going to put the nail in this beast called AGI. Working alone in one's private bedroom or lab is hardly conducive to creating the AGI or Singularity. A primitive cobbled apple computer in a garage by two visionaries of the past, perhaps, but this is a far different time for all of us. I would think that by putting the heads of like-minded people together and working toward a common goal should provide the best outcomes. No one is an island. ;) Thus this Forum...
Best of luck...
-
I did some work on common sense reasoning (https://artistdetective.wordpress.com/2016/06/15/how-to-teach-a-computer-common-sense), using "symbolic" knowledge representation, and while it ultimately requires a broader learning system to acquire more common sense rules, reasoning mechanisms work just fine in symbolic space.
In my opinion the "symbol grounding problem" is a red herring, it is indicative of people's inability to think of the world in other terms than their human sensory experiences. Words are just a type of information, just as are sound and sight, imprinted as electrical + and - signals on a glob of meat or a metal disk. The difference is only the richness of information: Words may certainly be considered a crude summary of reality with much detail lost. But the relationships, logics and interactions between concepts are the same regardless with which type of information they are represented. Words are already grounded through the people that speak them, or we would be spouting words in random order in random contexts.
Either way, one should not stare oneself blind on the words themselves, or risk overlooking the meaning found between them.
-
Yes Don! That is what I say too! Text has meaning data, tons. Humans invented each word for a reason, just like color. And the combinations of words as sequences and combinations of THOSE along with context is what can describe the sequences of the WORLD, that's what they are used for, they model the real world. Each word and phrase has a MEANING that can be said in text - like guarding means 'to fight aliens off from a kid'. All my knowledge is put into text actually, sure I know visual knowledge and sure enough my notes evoke back my visual thoughts I self discovered or learned online, but text is easy to write than draw and itt works enough to be a powerful AGI system is what I concluded. The only missing sensory in such AGI is...vision. Sound is covered already by text. As for touch, that I barely use, and vision covers it anyway. Of course more sensory = more ASI, and real world R&D will require real world sensories like touch, but again, text can cover its but for now.
-
^^^
just like humans invented color names...different languages have their 'needed' words they made up...phrases count as 'words' in a sense too......nodes
-
Further, colornames are words! Guys!, it's not just colornames humans invented, it's for all the word that were useful!!!
Ladder, steps, step, tread, rubber, side, rear, front, bottom, underneath, go, fetch, return, ball, cube, food, fries...man/woman
red, blue, green, yellow
-
Words are bite sized peices of experiance. They represent those thinkable bits of reality. Without the full human experiance of reality words would lose meaning. They would only be patterns of shape or sound. The vocabulary of an AI would be limited by the ammount of sensory data it can aquire and pair with symbols. Then if it has the sensory data, why not use it?
-
I also believe that no one person is going to put the nail in this beast called AGI. Working alone in one's private bedroom or lab is hardly conducive to creating the AGI or Singularity. A primitive cobbled apple computer in a garage by two visionaries of the past, perhaps, but this is a far different time for all of us.
What gives you this impression? AGI hasn't even begun as a field with any beginning principles, which means the field of AGI is wide open, awaiting new discoveries or ideas from anyone to get it started. Anyone can produce ideas without a computer, program, or a team of people. Only if someone gets hung up on the belief that AGI must be based on an existing type of hardware (e.g., computer, neural network) or existing type of software (e.g., algorithm, data structure) does that person become convinced that an individual is worthless for producing a breakthrough.
(p. 184)
Immunity Not Granted
In the marketplace, as untold millions have discovered, the
consequences of acting impulsively by joining the actions of the
crowd can be catastrophic. The so-called lemming effect (one
person following another over the edge of a financial cliff) has
fascinated students of human nature for centuries. The condi-
tion of your Self-Discipline Habit is severely tested when the
crowd is heading in one direction and your intellect and/or com-
mon sense tells you that it's the wrong direction. That's when
you must exercise the self-discipline to override the momentary
comfort of being in the mainstream. When you're tempted to
act impulsively in a situation like this, remind yourself that
the mainstream doesn't do very well over the long term. Civi-
lization progresses as a result of the actions of a few great minds;
the mainstream then simply goes along for the ride--and usu-
ally too late, at that.
Ringer, Robert J. 1990. Million Dollar Habits. New York, New York: Wynwood Press.
(p. 10)
The final article, "Thinking About the Brain," was written
by Francis Crick, the codiscoverer of the structure of DNA who
had by then turned his talents to studying the brain. Crick
argued that in spite of a steady accumulation of detailed knowl-
edge about the brain, how the brain worked was still a profound
mystery. Scientists usually don't write about what they don't
know, but Crick didn't care. He was like the boy pointing to the
emperor with no clothes. According to Crick, neuroscience was
a lot of data without a theory. His exact words were, "what is con-
spicuously lacking is a broad framework of ideas." To me this
was the British gentleman's way of saying, "We don't have a clue
how this thing works." It was true then, and it's still true today.
Hawkins, Jeff. 2004. On Intelligence. New York: Times Books.
-
I agree.
Lock doesn't follow the code/math/ANN way....I work on my own white-box. It does have ANN terminology, and very few math to it, and even lesser code to it, as code is just to code it. The efficiency of AGI matters yes, as it IS AGI, this is done in the 'idea', doing it in the code can be done by anyone later.
I agree, talk about what you don't know so you can realize what you do know.