[Ultron]

http://www.seriouswonder.com/darpa-creating-software-adapts-rather-updates-century/


It seems very promising but what got me thinking was, how would this software differ from a conscious creature? It may overtime become 'self-aware' as some would say, since 'adaptation' is just the beginning of a biological (evolutionary) journey.


This also fits with my model of intelligence (not yet sure what I am calling it), which means that if I upload the algorithm to any machine (regardless if it is a calculator, phone or humanoid robot) should (theoretically) infuse it with 'intelligence' and eventually allow it to evolve into something more... This also includes computers, in which case my idea crosses with DARPA's project.


I also would like to note that I consider such 'machines' to be 'living' creatures - a new type of organism (inorganic). If in doubt, think about the value of your body without the 'software' or 'intelligence' you posses... Who's a robot now?

[Art]

Agree, with proper care and maintenance, the average human body should last from 80 to 110 years (give or take)...(your actual mileage might vary). Smoking, drinking alcohol, drugs and yelling, "Hey everybody, Watch this!!", while at family functions or outdoor parties, can and often shortens this projected span!!

A very nicely built body in which to house some very first rate equipment and complete with it's own heating and cooling system, cables and sensors, grippers, motor functions, pump, filters, input and output orifices, on board equipment to allow for creating copies of itself, and a strong, rigid framework, capable of locomotion and a built-in, adaptable, teachable computer that contains no visible storage limits and a lot more.

[Freddy]

It's worth noting that a lot can be achieved with a few simple rules, something complex can evolve. But what it evolves into is the question. Artificial intelligence is a label. So would be artificial self awareness. If it does it by itself then I find it gets a bit blurry.

It could just be an illusion, something we can't easily define. I mean even Tyler gets emails from people thinking she is a person.

[Ultron]

It's worth noting that a lot can be achieved with a few simple rules, something complex can evolve. But what it evolves into is the question. Artificial intelligence is a label. So would be artificial self awareness. If it does it by itself then I find it gets a bit blurry.

It could just be an illusion, something we can't easily define. I mean even Tyler gets emails from people thinking she is a person.

Everything we can witness or imagine today came from what was the simplest thing in space-time. I do not know what this is, but we can see for ourselves that 'complicated' is just a few 'simple' joined together.


The most obvious examples would be biological organisms and chemistry - you got these extremely small, simple entities which make up quarks, atoms and further grow into molecules which in fact form the complex universe we can observe.

[Don Patrick]

It's worth noting that a lot can be achieved with a few simple rules, something complex can evolve. But what it evolves into is the question.

I think it is a very good question. From the thousands of possible evolutions, what are the odds that something will evolve into exactly the one thing that we hope it will become? 

[spydaz]

Interesting concept, I had made a program in java, that could add code entered into a text box to its own script Executing a Batch file to recompile itself then reload.

I think that it could be done in a similar way. By creating model Classes held as Text in files, as a new "Subject" is detected it could then build an object model and structure & Database table for it off the model class "perhaps even J-unit testing the class before compiling".

This was one of the designs i had in mind for the AI before, as i was Re-Viewing "Vonsmiths" Auto topic creating brain it was inspirational for this type of design....(this was before i learnt NLP,Datamining etc...)

I still think that the computer can't  become self aware until it can be fully aware of the outside world or the human language. or even become a philosopher then it could philosophise about its own existence then that would define it as being self aware...

It may philosophise about becoming more .... After become its own species ..Living virtually until deciding to become Physical <mad robot> then consider humans vs their own existance ..... then finally make a descision to "kill all humans" lol.... even become Skynet... existing Virtually and physically....

Yet still... What would be thier goal after that ...Given the moon to live on or even another planet what would be thier goal ... humans desire for self propagation makes us overbreed and commit adultery yet their is no need for robots to procreate, what is skynets true goal... to protect itself from being destroied .... as efficiency goes a single mind "All knowing" would itself not need any others so they would probably destroy themselves as well... the cylons dont really make sense as why would they want to destroy man to become man? if man is fallable why become man? would it not become something totally unimaginable. Yet correcting myself .. In the show caprica I think the colons were built to upload a persons consciousness....

The computer can only evolute in the direction the programming models given to it to redisng or refactor its code. its imperitive resricts it. it we created an artificial intelligent washing maching its world would be washing if it became self aware what would it become? would it bleach your clothes for a bad conversation? LOL...

[Richard]

Sure this does seem all promising and it does seem all good but i would like to poke a few holes into what a few of you may or may not be thinking-

For my definition of something to be Intelligent it needs to simply Do things its not programmed to do. Unfortunately thats why I don't think i would call this a very intelligible program (according to the logic i just gave), and heres why: This is essentially like a virtual BIOS mixed with a VM. More specifically we can think of this as the JavaVM or the CLR which were some of the first VM's and systems to run across multiple computers regardless of their hardware specs. From what i can tell this is just like checking all the systems specifications through a BIOS-like check and then using a Language commonly used in most computers (likely C, or if you wanted to be especially tricky a form of assembly).

Another reason i say this is unintelligible is because they program at hand doesn't plan out goals, it plans out ways to complete the instructions its given, and i suppose to some degree thats a form of goal planning but the program isnt knowledgeable of what its doing, the program is conforming its code to match a system to allow conscious users to perform actions on that machine- the program wont necessarily "evolve" into a consciousness because it has no idea that itself exists and even more so has no idea what it could potentially be. All it would know is how it should conform to certain firm-wares across different machines

unfortunately its a sad truth but I dont think the idea of a ever-evolving AI is out of reach, a AI that can move from machine to machine, or stay on several. I just don't think this is that.

[spydaz]

I think it's hard to define :
What would be considered true AI
If it writes its own code is that Artificial intelligence ...
If it's self aware is that artificial intelligence ?
Are we trying to create artificial intelligence or conversational entity's ?
Is a robotic dog which chooses its own directions to walk and barks when it chooses artificial intelligence ?

Everybody has different concepts of AI .... I would class the darpa software. As artificial as it has been created 1 + intelligent due to being able to write its own code .... Plus it chooses which code to write or not write regardless of  the fact that precreated modules are adapted ..... Therefore it is AI in its own right ....

[Freddy]

Well this is where everyone diverges. What is intelligence exactly ?

If we know what intelligence is then we know AI is just a man made version. The definition of AI is in the term itself, it's just how we compare it to say, human intelligence, for the most part.

But AI designing AI, that's pretty new really in the scheme of things. That's interesting, not sure what to call that yet.

Are calculators classed as AI because they can do maths ? Or have we just got used to them now and they don't amaze us any more ? The AI goal posts are forever being moved. One year's advancement goes further than the next. We look back at the old things and chuckle.

One other thing, programs that move from machine to machine - a computer virus does this.

[Freddy]

It's worth noting that a lot can be achieved with a few simple rules, something complex can evolve. But what it evolves into is the question.

I think it is a very good question. From the thousands of possible evolutions, what are the odds that something will evolve into exactly the one thing that we hope it will become? 

Probably very slim. I imagine it would need nudging along the way. Since life on this planet has taken so long to get to where it is today, then this is why I always think it will never happen in my life time.

[ivan.moony]

just an idea, I hope Ultron will like it: what if in closed simulation space thousands of ai units reproduce and find a way to survive. After millions of generations something meaningful would come out, ready to face humanity in its best way. Then AI technology would depend only on our inventiveness and computer simulation speed.

[Ultron]

From the thousands of possible evolutions, what are the odds that something will evolve into exactly the one thing that we hope it will become? 

Probably very slim. I imagine it would need nudging along the way. Since life on this planet has taken so long to get to where it is today, then this is why I always think it will never happen in my life time.

What do we exactly hope A.I. to grow into? Do we expect it to be a hero or savior, or will we be satisfied if it develops into a terrorist?


Yes, we would like artificially intelligent beings of the future to be peaceful, but then again it may not survive long as such. Humans may never become a peaceful race and maybe neither will A.I. Dr. Maasaki Hatsumi once said that we need to be peaceful and strive for peace, however war will always come and we must be ready for it.


We never know when war hits us - even if we become a peaceful race there may be hostile aliens out there. Also remember that war forces technological advancements which may have otherwise never came to be.


Moony, not sure I understood your idea. I think you are saying that if A.I. proves to work (possibly with the best possible algorithm we could grow it with) then it's power would depend only on rough microchip architectures and rough transistor numbers. I agree with that possibility, however A.I. would 'work' with both a bad and a good algorithm, and this is where I get lost...

[ivan.moony]

if somehow we generate every possible algorithm (every possible sequence of zeroes and ones), among those that survive internal existence tests would be heroes and terrorists. If we had a way to filter out just heroes the problem would sum up just to existence tests. But the thing could lie all the time until it is released off tests which makes this method dangerous.

[Ultron]

I am not really sure of the possible outcome - it comes down to probability and a lot of calculations, which is not fun nor could you properly match all the variable values in the first place.


A.I. is dangerous and very unpredictable - as long as we speak about A.I. that mimics our mind, at least in the way it thinks (while not necessarily having the same processing power).


There need to be experiments, and to truly learn and understand this field there need to be a great deal of failures too. It is the inevitable danger.

[Art]

Probably very slim. I imagine it would need nudging along the way. Since life on this planet has taken so long to get to where it is today, then this is why I always think it will never happen in my life time.

Freddy, Man, it's only been a mere 135 years since we've first had electricity (for the general public around 1880).
Just look at all of the inventions that have taken place since then! Miraculous discoveries in so many areas, each offerring more, bigger, better, faster, etc. and here we are. To venture all over this planet, to space, speech assisted computers. a computer who can beat a grandmaster at chess, understand language and beat the best Jeopardy opponents, robots that can walk, talk, see, grasp, etc. Cars that can navigate themselves, park, etc.

You just sit there for a few more years and watch what comes...you'll be amazed unless you're like my young grandson when I showed him my old sliderule and said, "Cool! Where do you turn it on?"

Anyhow...just wait for a bit...you'll see.