Digital Ecosystems

"I had a thought the other day, in regards to generating sentient AI.
Digital Ecosystem, designed to simulate billions of years of evolution. Essentially start with limited, instruction-based code that updates itself incrementally, and let these separate programs interact with each other.


Do you think this would be a viable course of action for Artificial Intelligence? If not, why not?


Regards,
Jonathan"


On the sliding scale of simulated evolution there's the purely organic abstractions and ones that rather closely resemble a collection of programs with an over-watch breathing down their neck to keep the population in check.

I remember messing around with this sort of thing in passing; Truly fascinating. For reference the program I believe I used back in the day was Tierra, an early 1990's approach at the latter; evolution of programs competing for CPU cycles and memory space.

The people who worked with Tierra found some rather interesting approaches at maximizing their chances of survival and breeding came about; For example there where parasites that used other programs replication code to breed, counter-parasites that tricked the parasites to create copies of said counter-parasite and other fascinating things.

Newer similar projects have the individual programs gaining bonuses for completing certain tasks and the like.

So what about other means, like the one alluded to in the quote?

What I have described above isn't really meant to go anywhere specific, it's more about the journey. Another way of using evolving programs that is used today is about the end.

In this approach you start with a blank section of code, or write a set of code that conforms to certain functionality requirements. Then throw the original version through the blender a couple of times. Test if the 'offspring' are any closer to a second more constrained set of functionality requirements, take the winner(s) then rinse and repeat.

And then there's the viability thing...

There is only one major limitation to all the above examples: the environment. Processor, Memory, and the fact that we are only processing abstractions of biologically similar systems.

With an unlimited amount of monkeys banging on typewriters processor power/memory, along with sufficiently complex methods of abstracting the biology in question (eg. enough potential complexity in the machine code of tiera or a good-enough facsimile of biological and/or neural function), we could very make a fully fledged AI - or in all likely hood -  an entire ecosystem of them with a bunch of monkey equivalents with oppose-able thumbs sitting at the top.


And how long did it take to make humans, out of how many planets?

That is not to say we can not learn a ton from these systems. They are invaluable when it comes to learning about biological and digital evolution. In the far flung future we may well see an AI arise from a long forgotten simulation left running.

First generation of AI however will not be from the evolutionary department, it will be from the engineering department. That is not to say adaptive techniques developed by both the humans making the simulatee's - and the simulatee's themselves may well be finding their way into future (or even the first) generations of AI.

Now after all that I have to say it: Lets just hope whatever machine that is running our own universe isn't running windows!

Moar Posts(?)

So the current line of posts is coming to it's end
Whats next? I could go quiet for a while 'till the next thing that must be brain-dumped appears.

Or I could take suggestions.

Enough from me, need more from you, my as of yet non-existent readers: what do you want me to do next?
Of note: I'm happy to do just about anything - I however won't touch religion for the foreseeable future

--end of line--

Artificial Intellegence Rights Part 4

Defining AI (continued) and defining the rights

It's easy to simply write up a feature list of an AI and say they get X rights for Y features. To use my last post as an example, each successive AI generation might get more rights because they are (in theory) capable of more, but also more responsibilities.

Obviously that won't because each generation may not be capable of more; just able to determine how you like your toast with more granular precision (he likes it 0.0001% darker on the butter side, but only on Sundays)

But a system like that is partway there, if is your right to have an internet connection, but do not have the capability to access the internet, whats the point? So the above needs to be used with the below.

In any case I'll stop listing features and rights and get down to the meat of it: 

How do we define those rights/Who qualifies for them?

WRONG TRAIN OF THOUGHT

Who defines those rights/How do we set the qualifications for them

Short answer: 

Democratically.

Long answer:

You begin with an (elected?) council of humans in Generations 00-01 (see last post if you want more info on generations) This is because any potential decision on the rights of even one AI (at least initially) will have major repercussions throughout the worlds of both human and AI. That and the niggling detail that the artificially inclined properly exist yet in a bra-burning sense.

Who are these people?

For the love of god not politicians, please, just don't go there. Aside from that i'd hope to see a mix of tech savvy philosophers. Bear in mind that is the ideal situation.

More likely your going to see major tech-developers/tech-gurus/tech-<insert relevant field here>, politicians representing the commercial sector (shudder) and/or military.

It ain't pretty but it's a start; and in the end, it is not about presenting a council without an agenda, its about presenting a council that is adaptable the the needs of the time and presents as many agendas as is feasible.

Why are they there/What are they doing?

They are there to regulate the industry to begin with, make sure that the cutting of the ribbon in front of the first true AI ain't' going to be a disaster, or if it is (and i'd put my money on it is), they cover it up. Besides, if they don't, someone else will certainly try.

Trust me, it's better for us in the long run. We want as little preconceived prejudices when the AI community as a whole starts philosophizing; Letting our fears get the best of us will set relations (and AI development as a whole) back YEARS if not damage them permanently.

Why? we oppress what we fear/hate. Can't escape from it, it's human nature at its simplest.

As someone wise once said: fear leads to hate, hate leads to violence, ??? leads to PROFIT  darkside.

...and yes I am well aware that it is:
"...Fear is the path to the dark side. Fear leads to anger. Anger leads to hate. Hate leads to suffering..." - A funny looking Muppet  YODA DAMNIT.

Where will they be going from there?

Well... that is a bit harder to predict, but easier to be naively optimistic about. Assuming all AI past the inevitable first crisis aren't lobotomized at creation you will eventually see some philosophizing; either by learned behavior or design.

These initial philosobots may or may not have anything useful to say on, for example, a council defining the rights for artificial intelligence's but those that do will help with the initial task of defining what an AI is. 

And for you sitting in your chair, even after reading my prior articles, saying "An AI is an AI, it's easy to define", I put this question to you: humans (outside of Africa) are from 5-10% neanderthal.

What is human?

And the council from there?

Just like any council it must constantly adapt, new AI's are elected to the council as old ones are shelved so to speak, definitions are redefined, humans phased out, post-humans - should they exist at all within the context of AI culture - phased in and we all go about out democratic life.

One last note, as I mentioned above, even an AI shouldn't be on term for forever, fresh ideas, oldies oldies being stuck in a rut (I know i'm stereotyping shush), young blood and all.

Or the council stagnate and fester like all councils do if left at it. After all for the most part "Governments are not voted in, they are voted out"

Post-Humanism

Since I brought post-humanism up in my last post I might as well put this footnote as well.

If we do not evolve with our technology our technology will leave us behind.

tl dr.

Artificial Intellegence Rights Part 3

Defining Artificial Intelligence

Time for a slight tangent for a brief and probably overplayed game of categorize that AI! Since most artificial intelligences will be radically different from each other a universal categorization system is rather hard, and yet necessary for the distribution of rights. More on that later though.

<music>We're talkin' 'bout artificial generations</music>
Since to exist a given artificial intelligence needs to be by definition be created, I can foresee an emergent timeline of said creation going something like this. Personally would like to see someone better qualified at this do one of these; this is as close to speculative fiction as you can get.

Generation 00 - And so it begins
The here and now; if it exists in the present (as of the writing of this article) it belongs here.

Generation 0.5 - The talking toaster test
In the near future chat bots and similar built solely to successfully pass the Turing test, I put this as gen .5 because as impressive as this is, it is merely the start.

Generation 1 - And so it REALLY begins
This generation will probably be more for proof of concept or specialized uses; testing the waters if you will. They will most likely be based off of gen .5/00 technologies and concepts such as the hypothesis generator .

This need necessarily be the case however. Biggest definer of a gen 1 AI is the implementation of heuristic learning methodologies and kicking of entirely new industries (or simply redefining a few old ones). I know the science of earthquakes for example would be revolutionized by a gen 1 AI.

Something else worth noting is that even though it will be based off of gen 0.5 tech, a generation 1 and above AI need not be able to speak; though it would most certainly help it in the whole proving it is intelligent (and thus an AI)

Generation 2 - The bubble
If it's happened once with humans, it's happened a million times, a bubble occurs, more money enters a given sector than is sustainable and (perhaps misguided?) innovation is rampant.  This is where I can see concepts such as self-modifying code bases being implemented in a practical manner.

Additionally this is where ill-conceived and poorly educated AI's with a lack of social skills and no work prospects leads to a brand new unemployed underclass, good job humanity. (I wish I could say that I am entirely joking when I say that)

Generation 2.5 - Pop
It isn't a bubble unless it pops; If there is going to be trouble it will probably be in gen 2-2.5 as progress moves faster than the society does. Heck, trouble caused by generation 2 will probably cause the bubble to pop.

What happens to all the half baked bastard children we've created though...?
And what will be the true backlash of the pop?

Generation 3 - Post-human is the new black
Signified by the designing and creation by an artificial intelligence in whole or part by another artificial intelligence. Post-humans also fall under this category because if you are not one by now  your officially outmoded as of several decades ago.

Also on post-humans, depending on the ratios of post-human to vanilla-human one or both will be  discriminating against the other. So you've shed your mortal shell, one or more humans, but now your all uptight know-it-all snobs who -still- haven't learnt from their own mistakes.

Generations 4+ going postal
At this point the sky is the limit and the definition of generations 4+ is probably best turned over those controlling it - our friendly new robot overlords/vacuum cleaners/unstoppable omnicidal doom machines, who knows this far in the future?

Artificial Intellegence Rights - Part2

Equal Rights for the ToasterKind?

So should we really give our toasters the same rights as we enjoy?

Short answer: maybe.

In truth it's a lot more complex than simply copy-pasting the bill of rights onto any machine that is able to tell us where to go (i'm looking at you, satnav). 

This is since, quite unlike humans, there are likely to be differences that count among st AI's. Your toaster, installed with a simple chat-bot (read: not smart at all) and able to learn how long to leave the toast in; probably shouldn't get more rights than you being polite to it every morning - and likewise it's only responsibility is to make your toast that perfect shade of brown.

A mainframe AI (read: smarter than the entire human race combined) would be subject to a few more rights than even we have; and notably more responsibilities/restrictions than we have. This is simply because it is capable of so much more than us squishy humans.

To put it into perspective, if humans where computers, we'd all be like commodore 64's; some with noisy fans and/or different colored machine cases; some able to use all their ram, and some are a little bit slow to boot up; but all essentially built to the same specifications - and the highest selling computer model around.

Now I understand that human beings have a lot more processing power than a mere commodore 64 - but suck it up, AI's are the eldritch abomination of the technological world; they hit the floor we might as well have that sorta processing power.

Equal rights movement for the Artificially Inclined? Nay.

Artificial Intellegence Rights - Part 1

Why even bother giving rights to a toaster?

Why indeed, there is a long precedence amongst humans that if two or more otherwise perfectly rational individuals don't quite see eye to eye everything form a bar-fight to bra burning to wars can break out. This is not to say that these thing won't happen for other reasons as well of course.

Humans being a generally selfish and stubborn species (I know peeps have their redeeming qualities, the only reason I have not jumped off this ride years ago really...) we will (by precedence) not give any more rights than we damn well have to until we are forced to.

This worked well and good when it was humans we where oppressing (see: all of history) - they simply fought tooth and nail through whichever means was available to them to right their situation.

This is not to downplay the countless battles won against oppression to this day - heck, many battles for human rights lay right around the corner, and it still won't be easy, just like it has not been easy in the past.

How does Selfish Preservation sound?

Now this is not why we should give rights to a toaster, but its a reason that may appeal to many people.Violence was one way, rights where clawed from an oppressors hands, politics but another - reform almost never came easy no matter the means. Now what happens when a vastly superior being of our own creation (read: just as bloody selfish and stubborn) wants it's share of the rights pie...?

This is where the future gets fuzzy, as there are a million ways an AI could be architectured and taught to do whatever it is it does. Will it rise up against us violently or will it do so in some other means using a technology that is inconceivable to our minds of today?

Will AI's of today's popular fiction ever exist when we can just put a post-human mind to work at the task?

Many many of the possibilities have been explored in books and movies - no real need to go over all them here.

The Future

Since this writing this blog is mostly in the future from the standpoint of this post start with, you guessed it, talking about this blog
From the title (The Past, The Present, The Future - As I see it) you can imagine I will be covering a broad range of topics, but anything specific I hear you ask?

Nay, I say: Everything.

If it's on my mind, it'll probably appear here, within reason that is; this isn't twitter after all. Fiction, Non-Fiction, restaurant reviews, you get the idea. But since this post is about the future in a vague sorta-kinda way, stick to that I shall.

I'll begin by a short disclaimer, I can predict the future. Ba-humbug you say. But wait don't close the page! You can to! It's easy! What will you be doing tomorrow? Bam done. You just predicted the future, congratulations. The future gets harder to predict with accuracy any length of time into the future - and frankly humans have a bad track record in the future predicting business.

The world is still here, Jesus has not come a second time, aliens have not enslaved us as foodstuffs and the USA did not get a bla- oh wait that one has come true, ah well, three outta four is a pretty bad track record anyway. This is not to that you can't predict the future, I am most certainly going to try a few (un?)educated guesses.

ps. Obama is a pretty swell guy, all things being equal, love his work. 'Cept for the medical marijuana crackdown though I don't blame him for that. That and probably a million other things I haven't heard of since I live in AUSTRALIA.

First!

I wrote an article later on musing on how far I had (not) gotten but there is no better place for it than the first page, so I am replacing the sad excuse of a first article and putting my reflection.


I feel stupid - the more I learn the less I know.

My articles have merely scratched the surface and I have made a decision to learn all I humanly can about AI, and to reinforce my learning I made this blog without the expectation that anybody would really care.

Well here I am reflecting on how little I have learned from the whole and yet how much more I know than what I knew. I guess this would be true of any subject - doesn't help me from feeling stupid as I look at the vast wealth of knowledge I am merely  glancing at.

And yet I cannot stop, I REFUSE to stop - I will not walk another path.

Anybody who wants to follow my forays into the world of artificial intelligence both technical and philosophical is welcome. This blog will not be entirely about AI. But so help me if it is not a major theme.

End of line.