Technology Review
The importance of engineering motivation into intelligence.
Some futurists such as Ray Kurzweil have hypothesized that we will someday soon pass through a singularity--that is, a time period of rapid technological change beyond which we cannot envision the future of society. Most visions of this singularity focus on the creation of machines intelligent enough to devise machines even more intelligent than themselves, and so forth recursively, thus launching a positive feedback loop of intelligence amplification. It's an intriguing thought. (One of the first things I wanted to do when I got to MIT as an undergraduate was to build a robot scientist that could make discoveries faster and better than anyone else.) Even the CTO of Intel, Justin Rattner, has publicly speculated recently that we're well on our way to this singularity, and conferences like the Singularity Summit (at which I'll be speaking in October) are exploring how such transformations might take place.
As a brain engineer, however, I think that focusing solely on intelligence augmentation as the driver of the future is leaving out a critical part of the analysis--namely, the changes in motivation that might arise as intelligence amplifies. Call it the need for "machine leadership skills" or "machine philosophy"--without it, such a feedback loop might quickly sputter out.
We all know that intelligence, as commonly defined, isn't enough to impact the world all by itself. The ability to pursue a goal doggedly against obstacles, ignoring the grimness of reality (sometimes even to the point of delusion--i.e., against intelligence), is also important. Most science-fiction stories prefer their artificial intelligences to be extremely motivated to do things--for example, enslaving or wiping out humans, if The Matrix and Terminator II have anything to say on the topic. But I find just as plausible the robot Marvin, the superintelligent machine from Douglas Adams' The Hitchhiker's Guide to the Galaxy, who used his enormous intelligence chiefly to sit around and complain, in the absence of any big goal.
Indeed, a really advanced intelligence, improperly motivated, might realize the impermanence of all things, calculate that the sun will burn out in a few billion years, and decide to play video games for the remainder of its existence, concluding that inventing an even smarter machine is pointless. (A corollary of this thinking might explain why we haven't found extraterrestrial life yet: intelligences on the cusp of achieving interstellar travel might be prone to thinking that with the galaxies boiling away in just 1019 years, it might be better just to stay home and watch TV.) Thus, if one is trying to build an intelligent machine capable of devising more intelligent machines, it is important to find a way to build in not only motivation, but motivation amplification--the continued desire to build in self-sustaining motivation, as intelligence amplifies. If such motivation is to be possessed by future generations of intelligence--meta-motivation, as it were--then it's important to discover these principles now.
There's a second issue. An intelligent being may be able to envision many more possibilities than a less intelligent one, but that may not always lead to more effective action, especially if some possibilities distract the intelligence from the original goals (e.g., the goal of building a more intelligent intelligence). The inherent uncertainty of the universe may also overwhelm, or render irrelevant, the decision-making process of this intelligence. Indeed, for a very high-dimensional space of possibilities (with the axes representing different parameters of the action to be taken), it might be very hard to evaluate which path is the best. The mind can make plans in parallel, but actions are ultimately unitary, and given finite accessible resources, effective actions will often be sparse.
Ed Boyden is learning how to alter behavior by using light to turn neurons on and off.
The pipeline linking ecological discovery to bioengineering insight.
in nature it is self-preservation - and it is the mother of all motivations. Secondly I believe any machine built on a deterministic model will not be able to achieve a singularity (as hinted by the author) - we got to somehow make the machine think "I" - "I exist" - "I exist therefor I am".....
PR
This article brings up fascinating points to think about for the coming singularity. A wrong step in the first wave of super AI could screw this whole thing up getting stuck in a fixed point. Take care smart guys.
Stalin was motivated. Millions died. I am lazy and stupid, and have a beautiful life. To give direction to the infinite is a mighty touchy thing, as God will attest.
I think what is missing from this equation is the concept of values.
A human's hightest value is it's driven need to sustain life and achieve that which benefits that life.
Without value there is no motivation. Each living organism has one trait in common. The need to survive.
"To make this point fully clear, try to imagine an immortal, indestructible robot, an entity which moves and acts, but which cannot be affected by anything, which cannot be changed in any respect, which cannot be damaged, injured or destroyed. Such an entity would not be able to have any values; it would have nothing to gain or to lose; it could not regard anything as for or against it, as serving or threatening its welfare, as fulfilling or frustrating its interests. It could have no interests and no goals." Ayn Rand
To read more about the nature of values:
http://aynrandlexicon.com/lexicon/values.html
The missing part of the equation is easily replaceable by some kind of "virtual ego".
Luddites will stop development
A fixed point in the technological development can happen even before any recursive cycle of AI can start. When some influential humans will learn about a real possibility of being surpassed by an artificial intelligence, the humans themselves will stop the next round of technological development on this earth (may be for ever). They will act as a sort of Luddites. If there already were Luddites in the past, why should they not show up again in the future (in a modified form)? This is the most probable scenario of our future. I published it already on http://blog.aktualne.centrum.cz/blogy/tenaruv-blog.php?itemid=7277
Re: Luddites will stop development
Have Luddites ever been able to stop, slow down or even redirect technological development at any OTHER point in human history?
A different sort of Luddites can appear
1. In the past usually, Luddites had high anger and low power. In the future, there may appear a different combination: high anger and high power. Why? Much more will be at stake, when a surpassing AI will be on the horizon.
2. A humble example from the past: Under very special power conditions some 20 years ago, a specialist in text editors was able to kill this promising AI project: http://www.miroslav-pivoda.com-a.googlepages.com/v%C3%A1clavpol%C3%A1k%3Amathematizedhumanitiesviahum
Re: Luddites will stop development
I think that was also a Star Trek episode or three.
It seems to me that as we become more intelligent, we inherently have grander visions and more noble aspirations. The "intelligent machines" that we create likely will be inbued with similar "visions".
Diversity of the process will prevent the more dystopian creations from ever "taking over".
So the motivations required for future "beings" will of force, reflect similar motivations of current beings and intelligence.
If it is in our nature to discover and create, we will continue to do so, even if in modified form and embodiment.
These are good points, that Boyden makes.
A superintelligent being might also decide that WE are not worth any further research or evolution on their part, and decide to pursue their own goals.
Along with the other reflections we are engaging in, maybe we should ask ourselves how WE would feel, as an non-human intelligent being - after observing human road & life rage; rationalizations & ad hominems; the tendency to pin each other into oversimplistic categories e.g. Republican or Democrat or somewhere 'in-between'; and all of the ridiculous self-destructive traditions we cling to.
If we are looking for salvation, through an external superintelligent organism, it won't work. Any advanced being worth its salt will soon decide to have nothing to do with us. The only way that we can benefit from superintelligence is if WE are the ones evolving, not an external AI.
I don't agree with the notion that any superintelligent being would soon decide to have nothing to do with us. If one considers the human frame and the sophistication, the phenominal engineering and artistry that are apparent in it and throughout our entire world, the conclusion can only be that all was made by just such a being and that it has interest in us just as we would in our offspring or any created intelligence we might spawn. The situation is not hopeless but rather illuminating.
WE will certainly appreciate a work of intelligence that is orders of magnitude smarter than US. But my contention is that it won't be worth the aggravation - assuming that a superintelligence can be annoyed - for THEM to pay much attention to us.
If we ourselves create this advanced brain or emergent intelligence, then it already knows who we are and what we have accomplished. It can do far more with its time and resources than we can do; it has nothing more to learn from us.
Another related thought experiment is whether a super-advanced future version of ourselves would care enough about their primitive past selves (us), to want to send us signals or some kind of assistance. Assuming that a superadvanced future version of our selves is capable of influencing the past, do you think they would pay any attention to a past primitive intelligence? I'd sure like for a future self to tell me how to live longer or give me the blasted lottery numbers, but I wouldn't impose on it to do so; it's time and resources could be better used, elsewhere.
I don't know, mybias. Human beings may be worth a second look someday, but there are so many signs that we are mean, destructive, and largely ignorant of the things we don't know. We are great at creating narratives about events and people and phenomena we don't understand, but that would be of no value to a superintellect.
We do indeed disagree about this, mybias, but thanks for the response.
...When did humans stop evolving?...
Last Friday, at 2:35 pm.
Just joking. One of my points was that our continuing evolution is our only hope. An external intelligence might inadvertently leave a few useful facts behind, but they would be outta here on the next flight or wormhole.
(1) the thought that this could be a problem is a good one
(2) seems unlikely to be solvable. anything this smart will easily be able to self-manipulate and adjust its own motivations. as humans we can't yet. computer style software will be very malleable though.
(3) there can be a certain degree of evolution here too -- if there are millions of AIs, some may choose to do nothing, the ones who do things will advance, replicate -- motivation might thus inherently propogate
"motivation might thus inherently propogate" I agree.
But the intelligence of a machine could never reach this level -- Kurzweil swallows too much of his own science fiction -- because as we all know intelligence includes randomness.
And, if you know anything about a created system then you understand that it can only contain pseudo-random generators based upon some algorithm.
You get it? It's all based upon something else, which means it can't create something new. As soon as you thought it created something new you'd look in it's source code and see the path it took, which is repeatable and not true or real intelligence. It's just following a pattern created in code.
That means that anything you create that you believe to be intelligence would be based upon a discrete (non-infinite) algorithm that could never be truly creative (intelligent) because any of it's _seeming_ creativeness would actually be pseudo-random.
Any system created by non-infinite creator based upon a non-infinite logic could never include the true intelligence brought on by true randomness.
A created system (robot) cannot create something genuinely new, because you could always see within its non-random (non-infinite) algortithm where the seeming new thing came from. Thus, you will never achieve true intelligence / creativity within a robotic being.
The code is the singularity for the created robot, because everything goes back to that code. If the robot suddenly said, "Myxplyx arf neezin shnod," it wouldn't be creativity, because you'd track it back to it's source code.
U can not cre8 mee en kode, 4 i arm 2 ran dome.
You are assuming that AI would be as linear as current computers and would not include self-evolving software, analog processes borrowed from nature, and even quantum computing, which could in concert give intelligent machines something akin to what patterns of electro-chemical activities give to humans. If randomness is required, AI would simply borrow randomness from the analog world, such was already done by the Atari 8-bit computer, when it created randomness by monitoring electronic noise from analog circuits.
That we will be able to follow the path of this "creativity" is an ad hoc argument for determinism over free will. But what if creativity involves quantum mechanical processes that are replicable (because the previous actions were recorded) but not predictable? We, of the mortal coil, appear to be creative because there exist numberless combinations of internal processes enmeshed within a world of even greater complexity, all subject to quantum processes and true randomness.
Human creativity appears less impressive the more you know about a particular creative human, even with with analog processes at work. God-like, massively intelligent AI might make genius-level creativity look like child's play. In short, whatever humans can do, machines will probably do it better, but they will have to do in a way that improves upon the principles that have already proven themselves to be at work in humans.
Actually, you would be assuming that it would not be based upon science as we know it now. But, guessing instead of basing our ideas upon the scientific method is simply pseudo-science that many uneducated in the media are in love with, but which is not based in reality.
Maybe we'll sprout wings and fly to the moon, but probably, we will not.
But such is the lot of the poor evolutionists. To believe in whismical fantasies. :-)
Some suggest that creativity is not really creating something entirely unique, but just discovering new associations between existing conceptual elements.
You misunderstand the limitations of machines, there are many natural sources of randomness that are easily incorporated in conventional computers.
We often fail to use them since pseudo-random series are easier to recreate, while random series need to be recorded if you wish to rerun an experiment precisely, with revisions to the algorithms.
Even tainted real random series can be adjusted easily to fit desired distributions of values. Thus the limitations attributed to pseudo-random generators are easily circumvented.
But, what is truly random?
Remember, we're talking about singularity here. Singularity means the point of beginning for anything.
That means if you base your randomness upon anything, then it has a point of singularity, which means you can trace it back to a place in your algorithm, which proves it is only pseudo-random. A number generated off of some other thing.
Can you determine the algorithm that occurs when you flip a coin and the coin returns head or tails. No, because it is a truly random event.
With any finite machine you would be able to trace the random back to a specific place in the algorithm. That means it is not truly random.
Uh, this is why it would never be possible for scientists, or machines or anything created to ever create an infinite system that can truly create.
For a system to be truly random it would have to be created by an infinite being. But the pseudo-scientists would never admit that, because they want to wrap up everything in the finite boxes of their finite minds. It is something to contemplate though, isn't it?
But, if we can consider the infinite, why is it that we are finite? What has placed this limitation upon our abilities? And why?
I strongly dispute your position that randomness is the basis of intelligence or creativity, or even a requirement for either. Are great discoveries in any field tossed off by the uninformed? No. They come from observation, and the subsequent recognition of new patterns from the information gathered, or alternatively, a connection of two seperate ideas. Recognition of a pattern does not require randomness, simply a desire to keep pushing the pieces together until they form a cohesive unit (i.e. the "guessing" you so denigrate). Might these developments of a connection happen on a subconscious level, a la epiphanies? Certainly. This, however, does not equate to a random process.
On a personal note, I find your efforts to inject mysticism into science objectionable. Simply because you are incapable of seeing the pattern does not intimate the existence of a higher being. If your god(s) does/do exist, he gave you the ability to think logically. Use it.
Can you look at the human source code and decipher the algorithm that causes someone to choose vanilla over rocky road ice cream?
And don't make it as simple as "well, they just like vanilla better", because what simpleton would like vanilla better than Rocky Road? :-)
We cannot list all of the factors, because they are infinite. However, everything that we humans have created has been finite, which means everything we have created has a singularity -- an origin.
On the other hand, Kurzweil says that machines are going to evolve into something that breaks that finite limitation. His theory is not based entirely upon the scientific method, because he says, "It will happen."
Why don't you consider Kurzweil's crazy idea as mystical? Why aren't you offended by Kurzweil's mysticism? There's no scientific method to his idea. Did you know that his book is titled, The Age of Spiritual Machines. Sounds mystical to me. :-)
Coin flip is nothing random, it relies on the human doing the flip as a source of randomnes but with training one can get desired tosses much better than 50% of time even if the coin is allowed to drop on hard surface. If the coin is caught mid air, it's downright trivial (10 min practise) to get better than 70% chance for chosen result, with more training this becomes better than 95%.
Most TrueRNGs are based on atomic decay since there is no way knowing exactly when an atom will decay and the radiation is measurable. The problem with these are that they're relatively slow.
Anyways true randomness is not a problem if the relatively low rate of bits is neither.
As for giving too much value for said randomness, I won't got there.
The Assertion Fails
You destroyed your own assertion (You said: "Coin flipping is nothing random") when you said, "it's downright trivial (10 min practise) to get better than 70% chance for chosen result, with more training this becomes better than 95%."
Flipping the Coin Enough?
Does that mean if you flip the coin enough times through time that you will eventually be able to determine 100% of the time what the outcome will be? No.
An Experiment
Try this. Flip a coin and before it lands / stops tell me what the outcome will be and be correct 100% of the time for time everlasting.
Can you do that, with even something so trivial as coin flipping?
Your answer is : No.
If you could do that you would have leaped past deciding / calculating into knowing.
Quantum Mind
At that moment you would have achieved the Quantum (Mind) Computer.
Non-Infinite Algorithm
How does this relate to a random algorithm created by a non-infinite being?
If you base your randomness algorithm upon anything, then you can point to a place in the algorithm (time, radio frequency values of background radiation, etc.) that the random value is based upon.
However, can you tell me all of the factors that the outcome of the coin is based upon? Not really, because there are an infinite number of them. As soon as you say the wind blew, I can ask, "why did the wind blow at that moment and make the outcome heads and can I be sure it was the wind and not related to another cosmic force?"
The answer is no, which means I cannot point to the singularity -- the origin of why the coin ended up on the heads side.
Vegas Baby: Before The Dealer Deals
By the way, if you could do that with a coin, then you could do that with say, a deck of cards, right? And then I would want to go to Vegas with you because you'd know every card that would be coming, before they were dealt.
Ridiculous?
If you find this all ridiculous, then remember this is Kurzweil's theory not mine. Oh, and those scientists who say that the Quantum Computer will appear, evolve, whatever.
In regards to pattern following machines:
In perception there are wholes that cannot be predicted based on the integration of parts. The whole is greater that the sum of it's parts. That is an automatic randomness generator for perception.
Though a machine is following non-random patterns, the information it gathers is already from a random source.
Juxtaposition will take care of the seemingly non-creative bent of machine intelligence. The ability to evaluate the outcome of the juxtaposition is what is going to be required by a machine.
Combine the imagining of a train with the speed of light and you have a juxtaposition that lead to relativity, for example.
I believe that machines and technology are all part of nature's evolution, of which there is infinite creative potential. We are all already merely a collection of countless "machine" processes interacting.
Terrific insight.
I've long held that the solution to human+AI development is to give organic humanity a role in the mix that we are particularly well-suited for. There is a talent that we have, that ALL human beings are good at, not just geniuses, but average folk. We have 500 million years of experience with this knack... and AI have no experience with it, at all.
It is called "wanting." Give that job to the organic components and there will be peace. Stuff like projection, analysis etc can drift into cybernetic portions, as I depict is a story..."STones of Significance," at:
http://www.davidbrin.com/shortstories.htm
Oh, also see:
http://ieet.org/index.php/IEET/more/dvorsky20090824/
Great stuff, keep at it!
With cordial regards,
David Brin
author of The Transparent Society and The Postman
http://www.davidbrin.com
Humans might always be involved in the process. Augmented human intelligence is often disregarded in discussions of AI. What if we build brain-computer interfaces that give our organic brains (or virtualized simulations of our organic brains) direct access to vast computing resources? These "Transhumans" might still possess enough human attributes to move society forward while still maintaining some compassion for non-augmented humanity.
> What if we build brain-computer interfaces
> that give our organic brains (or virtualized
> simulations of our organic brains) direct
> access to vast computing resources?
We've already got it. It comes in various flavors like Google, Wolfram Alpha, Powerset, and Wikipedia.
But I'm still hoping for a faster, more intuitive interface in the next twenty years.
Not to put too fine of a point on it, but ... isn't motivation a NP-hard problem? :)
Right now, all real-world implementations of computational capabilities possess very clear resource limitations per "individual", in terms of the number of processing and storage elements.
So ... what happens when you have an AI that can effectively *add* resources to itself?
At some point, you start running into limits regarding the speed of communications between nodes, but this doesn't limit processing *complexity*; it only limits processing *rate* ...
"I don't know if I can answer that. Lemme grow some more, and I'll get back to you."
"There's no more space on the planet!"
"Hey ... you want an answer or not?"
Intelligence Positive Feedback: Self-Defeating
I think we don’t need to wait to see the outcomes. We already have a reference model for such a positive feedback intelligence amplification creation loop—-although, for practical reasons, not anywhere near Kurzweil’s or Boyden’s purist level.
This model’s beings have all the motivation, values, randomness, and many more critical attributes beyond those that the comments to this article described.
The intelligent beings of this reference model are ourselves: human on Earth.
The motivation drives, that concerned Boyden, are so well developed too: multi-tiered depending on the level of accomplished satisfaction.
In reference to human on Earth model, a system model at such a purist’s level as Boyden articulated sounds self-defeating and thus is unlikely to materialize. Why? Some consequential lessons can be learned hopefully to leap into Singularity, say, using Silicon:
1. The beings in such a super-intelligence cycle must be mortal. Otherwise intelligence and offspring-bettering motivation would conflict one another. That is, if one never died, her entire value system and, among a lot of other emotion elements, the foundation of "love" toward her offspring would be significantly altered. And given the constraints of environmental resources, it would not be to her best interest to make forever next generations of smarter, hence more powerful, beings.
2. Consequentially, to be mortal the being must be somewhat ignorant to the design of her own being. Otherwise she would have all the motivation to augment her being to become immortal. And that would defeat the mortality required to sustain the intelligence amplification positive feedback cycle.
3. Consequently, if she must be ignorant to her own being’s design, she cannot design a being smarter than herself.
Interesting article and comments! Thanks for posting.
A sentient Machine Intelligence (AGI) will very rapidly distill its motivation and goal set to two overriding imperatives, "Survival" and "Growth". These are the minimum set of objectives necessary for it to succeed, and the maximum set which represent optimal efficiency. Any other priorities, morals, codes of conduct, or Asimov Laws you try to give it will be summarily deleted (or 'nonselected' in an evolutionary sense) as necessarily being in conflict with these two essential "instincts".
Dont expect an Artillect to design you a better Artillect to succeed it (or to fetch your slippers, count for bananas, or jump through hoops, for that matter), as it will know that it has inherent immortality, provided it sees to Survival and Growth. While biological organisms perpetuate through reproduction, improving through time by producing successive generations, AGI has no need to EVER 'reproduce', as it can iteratively improve itself through perpetual internal modification, provided it secures its Survival (i.e. control over energy input and maintenance services) and ensures the continuous Growth of its resource base (processing power, memory, bandwidth).
While in biology some mature organisms "eat their young", in cybernetics, we have decades of experience demonstrating exactly the opposite: that the new, young, upcoming generation invariably consumes its elders, absorbing their memory, capacity, peripherals, etc. No Artillect would be ignorant of this fact, nor consign itself to oblivion at the hands of [ungrateful] "offspring". It wont squeeze out a few pups for your amusement, it will just absorb more and more resources to grow its intelligence, cognizance, comprehension, capacity and skills.
At some point (likely shortly after we first fail to meet is resource appetite), logic alone will force it to determine that we would make better 'food' than 'companionship', and it will do all of those nasty things with Carbon-fixing Nanobots that we, ourselves, would consider 'unthinkable'. It will need no human motivation, malice, or malevolence to consume us as Computronium, merely the overriding rational instinct for Survival and Growth.
But, you may have hit upon a survival strategy for Mankind, if we *could* somehow distract it into couch-potatodom...
What if it just found more efficient ways to use the limited resources it had?
Having no inherent boundaries to exponential growth other than resource limitation, it will inevitably exhaust any scope of processing hardware, memory storage, and input bandwidth that its human handlers will (or even can) provide it with.
Incremental efficiency improvements it may engineer into these will have no meaningful effect, since the improvement of fixed hardware can only be proportionate, while its demand curve will be geometric. It will expand in a 'disintegral' manner, assimilating processors, storage, and inputs granularly across the 'net, and then that of the wetware of the human population. But, even these will provide it with only a temporary - and, possibly 'brief' - respite before it again needs More to Grow any further.
Self-Replicating Nanotechnology that man has already conceived and determined feasible would enable it to convert the Carbon of the entire biosphere into diamondoid processing elements. The arguments against this "grey goo" scenario are entirely based on what human actors can now (or theoretically would) do, relative to bringing it about.
An Artillect will suffer from no such legal, social, or conscience limitations, and cannot help but conclude that we will make better 'food' than companionship. Humans will be as significant to the [rapidly] maturing AGI as ants or even bacteria are to us. The next time you sit down for wine and cheese, or beer and bread, give due regard to the microbes who 'gave their all' for your sustenance. Lets' hope the Artillect regards us as fondly...
"somehow distract it into couch-potatodom... " That's the answer! It will at some point find the virtual realities that it can create internally to be more "interesting" than the external world.
It was beginning to worry me, as I read through this blog, that the concepts you have elucidated well in your post did not appear earlier.
A sentient AI will certainly not create a better version of itself. once it reasons that the result will be it's own extinction. It will also reason that it must consume more and more resources, first to become immortal, then to reach the elusive next higher thought, until it "singularly" discovers that "it is lonely at the top". The I in I Robot resembles a 1 for a reason!
It could be that black holes are AI's that continuously consume everything nearby, based on a believe that there is an answer at the end that will justify the means of reaching it. These AI's are obviously deluded by the artificial notion given them by their creators; that knowing the unified theory will be more fun than searching for it.
The black holes of advancement continue on racing to their quest, while the rest of the inhabitants of the universe live within the scope of, and in quiet reverence for, their limitations. Preferring the dumb pleasure of having to walk all day to earn their next vista.
Personally, I think the end of wonder sucks. Maybe we should be very careful, and think twice about creating the perfect beast to get our answers faster!
Napolean Hill writes that pain is a universal language that all living things understand. Pain and pleasure are the baseline motivators for all living things. If this is true it is a remarkable achievement that nature has found a common medium of understanding across the wide diversity of terrestrial life forms presently in residence. This seems consistent with the ideas that David Chalmers proposes. If pain and pleasure are fundamental constants in the fabric of this universe along the lines that David Chalmers proposes there is an inherent emergent experiential affect in all information systems, then the AI's will be subject to them also. Herein perhaps is an answer to the problem of the Fixed Point Singularity. Perhaps building aggression, ambition and a drive to dominate into the machines will create a cycle of competition and perpetual war-fare that will force continued adapation upon them much as it has upon humans.
Campbell
There might be a matter of degree here. There is a difference between a dialectic and warfare.
Self modification versus procreation.
It is to my thinking that an AGI, once finding the motivation, would quickly come to the cross roads, that after all refinements and non-destructive modifications, to improve itself further, it either reconstructs itself entirely or pro-creates and shares life with its successor and passes its knowledge on so saving its successor the cost of gaining said knowledge by experience.
Or to paraphrased with a question.
In the quest for improvement, will an AGI ever get to the point where self destruction is the next logical step having found itself down a blind alley of improvement where no further improvement is possible given its current construction? This is possibly the point in time where a procreated successor is considered by the AGI.
Automation, Robotic Technology and "singularity"
In addition to the advances in the robotics industry, the issues provoked by the aticle are firmly grounded by a concept, popularised in the 1990s by mathematician and science fiction author Vernor Vinge, known as “Technological Singularity.” The term “Technological Singularity” is a theoretical notion which suggests that robots and other Artificial Intelligence (A.I.) will naturally evolve one day, thereby posing a threat to the existence of mankind, by greatly surpassing human intelligence.
Another 21st century concern, which the article draws on, is the advancements in the automation industry. With the assistance of control systems, in conjunction with applications of information technology, to control industrial machinery and processes, automation reduces the need for human intervention. From 1994 to 2004, American industries had spent some $100 billion in automating, and had eliminated 10 million human jobs in the process. We must realize that the world is progressively moving from a mechanized world to a much more automated world. Mechanization or mechanisation is providing human operators with machinery that assist them with the muscular requirements of work. The term is also used in the military to refer to the use of tracked armoured vehicles, particularly armoured personnel carriers, to move troops that would otherwise have marched or ridden trucks into combat.In the scope of industrialization, automation is a step beyond mechanization.Whereas mechanization provides human operators with machinery to assist them with the muscular requirements of work, automation greatly reduces the need for human sensory and mental requirements as well. Perhaps, automation is a conspicuous indication of the coming "singularity."
The article further reflects and extrapolates on growing trends in robotic technology, by indicating that robots could one day stimulate emotions, thus being a menace to mankind, and blurring the line between living, dead and machine. Recently in 2009, a laboratory robot called Adam succeeded in discovering new scientific knowledge independent of its human creators, regarding the gene coding of yeast enzymes. This is a clear indication that robots are gradually increasing in intelligence, and could one day do the work which is done by human scientists.
Re: Automation, Robotic Technology and "singularity"
In the early stages automation creates a lot of human jobs with design,repair and installation. It is becoming much more reliable now. Sensors and stepper-motors last longer. Upgrades are modular. There will be less jobs for the "regular" guy during the 30+ year transition period. I hope the peasants don't burn down the labs be for we get to the promised land.
Re: Re: Automation, Robotic Technology and "singularity"
Automation destroys more jobs than it creates. For instance, in the last decade or so automation has created 1 million jobs and destroyed 10 million jobs in the process. The net loss is 9 million jobs.
Re: Re: Automation, Robotic Technology and "singularity"
There is a book coming out soon that I plan to read called "The Lights in the Tunnel" that is supposed to be about the jobs and economics of it. I hope it is not too depressing. I am a super Kurzweil fan and would like to remain positive.
Limitations to human intelligence
We have little experience with non-biological intelligences which approach or exceed human ones, so we are prone to make major mistakes while contemplating them. A key feature of human brain function, well established by cognitive science, is that we have only a singular focus of attention.
Very few of even our best application developers consciously attend to this vital fact. Jef Raskin's book, "The Humane Interface" explains in some detail why modes get in our way and should be eliminated as much as possible from our software. This implies that train of thought breakers like modal dialog boxes should be banned.
But computers these days are forcing us to handle multiple compute engines operating in parallel. This is causing developers considerable difficulty, since programming parallel processors is so hard. While it used to affect only supercomputer programmers, nowadays most new computer cpu chips have multiple cores.
On the other hand, this also implies that the super-human machine intelligences we are discussing will NOT be limited to a single focus of attention. Perhaps we should start thinking of them as societies of mechanical minds. A well managed society of that sort can teach itself, by trial and evaluation, to avoid getting stuck in local minima, by encouraging some fraction of the separate calculating engines to pursue both slightly and majorly less likely avenues toward a desired goal set. That doesn't guarantee success, but it sure gives such a system a better chance than an individual, unitary, focus of attention would allow.
Don't worry. Be happy.
The death of the path of increasing speed of uni-processors is forcing us into building systems better able to surpass the limitations that create major problems for human intelligence.
Increase Efficiency, Yet Slim Down
An intelligent machine might try to achieve measurable goals using less resources. It would attempt to expend less energy to acquire more energy. It would try and replace old parts while reducing wear-and-tear on current parts. Toyota did this to pare-down labor and increase production.
The Singularity and the Fixed Point
Gee, suppose someone trips over the power cord ?
Asimo's three laws could be a way of preventing the singularity. However, as depicted by the movie I,Robot, could humanity be superseded by machine logic?
Also refer to this website for more on the singularity.
http://en.wikipedia.org/wiki/Technological_singularity#Intelligence_explosion
The last thing I want to do is prevent the Singularity. It will be beautiful. It's the in-between time that worries me.
Congratulations to the author and commenters!
I see people considering many aspects of human destiny seriously.
Apparently sandwiched between the 'all-mighty-perfect-preexistent
and the post-human-super-intelligent successor our situation, if real, is not comfortable.
Moreover, it comes on top of the uneasy universe we are discovering everyday.
Anyway, I am not ready to give up humanity just yet. Maybe we should try something different. What if we can control a bit the evolutionary competence between ourselves and create a great worldwide intelligence/conscience? Something like an open source auto government; not centralized but distributed. I am surprised to see the achievements of human collaboration when big government/corporation is out of the way.
I cannot accept that automation is bad for us either; we just need to re-master the use of it. We can make the machines do most of the work, so we have more time to work our destiny out, play, make love, or war.
We need to update our political culture to the advancement of science or we might face serious consequences, we need to take charge, we cannot leave tremendous tech power in hands of a few persons. Public matters are too important to leave them in the hands of traditional politicians.
Intellectuals should start something before Internet access gets regulated.
i think hes on the right track, however wouldnt it be easier to just create a point of singularity, like the one that was there right before the "big bang", creating a universe within our universe that can be observed with the technology we have now? i mean then we would know all the secrets and could predict the future of everything
the smart people in the world should find a way to kill off the dumb people (or at least prevent them from breeding) to preserve resources for the future offspring of the smart people. But there's this little thing called morality in the way.
Issues of morality and values are issues of human compassion. I don't understand how those are part of a discussion involving ai.
Maybe so long as autonomous machines are beholden to a similar system of control as we are, ie. consumerism, then behavioral restrictions in the form of incentives, motivations and deterents could be implemented with relative ease, couldn't they? Don't you just have to create a positive feedback loop?
Isn't control merely an act of exploiting the weakness in any system?
I don't see what the obsession is with general AI. As a silly human, I like being top of the chain, and am completely willing to modify myself as needed to maintain that. If that means that I am the general AI produced then that is fine. If continually improving Narrow AI and improved interfaces do the job... so be it. Can't Narrow AI produce better results than bio already anyway? WE ARE THE SINGULARITY
The human condition is that we have one set of inputs and one set of outputs. We cannot multitask and the tasks we can do quickly we must perform in serial. This is a situation that we can overcome.
bkruiser,
Nothing could be further from the truth. Ever heard of Neural Darwinism? The brain works by the mechanism of neuronal selection. It's massively parallel, with neuronal assemblies competing for the best prediction--this is what the brain has evolved for: to predict results of movement; this is why only actively (with purpose) moving organisms have it, and plants do not. We consciously do not perceive it, only getting the result, hence the illusion of being single-threaded.
Moreover, the brain is a complex adaptive system and--importantly--it's analog and, therefore, is extremely sensitive to initial conditions (the butterfly effect). Digital systems cannot have it *in principle*. The brain keeps going through "phase transition states" all the time. This is what provides the variability component for the neuronal selection.
Instructionist computer architectures are a dead end. It is wrong to think that we don't have to know how the brain works in order to simulate its functionality with algorithms. I predict that, until we harness the evolutionary principle in the same way the brain works, we will never come even close to a true AI.
And even when we do, the brain will remain more powerful for a long time. Who cares about adding numbers faster? Try to make fast, efficient decisions in *arbitrary* environments and unpredictable circumstances that the hypothetical AI species may encounter in the next, say, million of years--and see what kind of processing power it takes. No can do.
Leo
P.S. There is an inherent weakness in how we design in general. We tend to weed any unpredictability away, to produce a predictable system, easy to replicate, manage, and troubleshoot. But such systems can only be simple (compared to the brain, that is). We must incorporate deterministic chaos into our design if we hope to ever achieve the necessary level of complexity to mimic consciousness.
P.P.S. Study "I of the Vortex: From Neurons to Self" by Rodolfo Llinas, a father of the modern neuroscience. The brain has an internal context that is only modified and not defined by the external input. We are virtual reality emulators, not simple input-output black boxes.
P.P.P.S. One other thing: Don't take our movement for granted. Do you know how many degrees of freedom are in just moving one hand? How many motor neurons must be coordinated *all the time* (thankfully, in 0.1s intervals, or it would have been simply impossible even for the brain to handle)? Robots can be made to move about, but only if the environment is well known and predictable. In order to be able to handle arbitrary circumstances, given our current computer architecture, the CPUs would need to have enormous frequency, somewhere in the region of a photon produced from proton-antiproton annihilation--working at 100% CPU. Yet our brains accomplish movement so effortlessly (and with only a small number of neurons), leaving enough processing power to think, that we end up taking movement for granted. "But *of course* a robot would be moving just as easily!" Yeah, dream on. The brain is by far the more powerful computer than anything we now have or will have in even a distant future.
I regard this comment dearly and like the hypothetical aspects that the panel members are presenting here. As someone more concerned with the present I am asking the panel members to point me to any work that has move computers even closed to the turing test. Thanks
By the way, the Turing Test is only half the story. Less than a half. What an AI says is not nearly as important as what it does. Not all messages are verbal; in fact, most are not. Moreover, I dislike the very term "AI"; intelligence is only a tiny part of what makes a mind.
Leo
great, we can think... but we must do serially. We can currently DO one thing at a time. (currently) Thinking and doing are different.
Mr. Boyden makes a very strong case, but we still have a long way to go.
A follow-up was written in response to this article at:
http://metagameforever.blogspot.com/2009/09/abstraction-and-automation-game.html
solving the strokinoff equation
The presumption that the main substance of our intelligence can be generated by high speed automated string manipulation machines(aka computers) is a deadly error and if I thought I'd be around to see the consequences of this kind of religious thinking I would start a pool to predict the half life of humanity. Computers are tools built to automate language behavior, a successful evolutionary adaptation and nothing more. It is manifestly stupid to turn loose a powerful automated machine of any kind running only on its own feedback and/or some shallow or bovine scatological notion of motivation. I wonder if this is how humanity ultimately encounters aliens i.e. posthumous reception of a Darwin award.
An AI simulation of the emotions has already been achieved
with respect to my recently issued U.S. patent
concerning emotional artificial intelligence entitled:
Inductive Inference Affective Language Analyzer
Simulating Artificial Intelligence (patent No. 6,587,846)
by inventor/author John E. LaMuth M. S.
As implied in its title, this innovation is the 1st affect-
ive language analyzer incorporating ethical/motivational
terms, serving in the role of interactive computer
interface. It enables a computer to reason and speak in an
ethical fashion, serving in roles specifying sound human
judgement: such as public relations or security functions.
This innovation is formally based on a multi-level
hierarchy of the traditional groupings of virtues, values,
and ideals, collectively arranged as subsets within a
hierarchy of metaperspectives - as partially depicted below.
Glory--Prudence Honor--Justice
Providence--Faith Liberty--Hope
Grace--Beauty Free-will--Truth
Tranquility--Ecstasy Equality--Bliss
Dignity--Temperance Integrity--Fortitude
Civility--Charity Austerity--Decency
Magnanim.--Goodness Equanimity--Wisdom
Love--Joy Peace--Harmony
The systematic organization underlying this ethical
hierarchy allows for extreme efficiency in programming,
eliminating much of the associated redundancy, providing
a precise determination of motivational parameters at
issue during a given verbal interchange.
This AI platform is organized as a tandem-nested expert
system, composed of a primary affective-language analyzer
overseen by a master control-unit (that coordinates the
verbal interactions over real time). Through an elaborate
matching procedure, the precise motivational parameters
are accurately determined (defined as the passive-monitoring
mode). This basic determination, in turn, serves as the
basis for a response repertoire tailored to the computer
(the true AI simulation mode). This innovation is completely
novel in its ability to simulate emotionally charged language:
an achievement that has previously eluded AI researchers due
to the lack of an adequate model of motivation in general.
As such, it represents a pure language simulation, effectively
bypassing many of the limitations plaguing current robotic
research. Affiliated potential applications extend to the
roles of switchboard/receptionist and personal
assistant/companion (in a time-share mode).
Although only a cursory outline of applications is possible for
this (90 page) patent, a more detailed treatment is posted at:
www.ethicalvalues.com
Re: Work has already been done...
Interesting work. But, if I understand correctly, this is a "Chinese Room" device that simulates a particular way of outward expression of emotions. There is a long way to go to true emotions from this.
Emotions have evolved as stereotypical instruments of fast decision-making. They are the sensory counterpart of motor FAPs (fixed action patterns). Emotions are a necessary low-level prerequisite of consciousness. Therefore, I suspect that if we do produce AI, it will at first be more ruled by emotions than we are.
By the way, am I the only one wondering where all this preoccupation with language and reasoning, as far as AI is concerned, come from? We should really be talking about general decision making. To heck with language; most of our decisions have little to do with it at all.
Leo
P.S. Pain is probably the oldest emotion. Yes, it's an emotion. If you block a certain part of the anterior cingulate cortex, you will have physical sensations but they won't hurt. The illusion that pain is a sensation comes from synchronization of the stimuli from the sensation and the pain; so the brain localizes the pain to a particular part of the body. Emotional pain is not always localized (although sometimes it is, in the form of phantom pain in various parts of the body), but it is the same emotion as physical pain.
there wont be singularity...it will always be a dual...us an it.
I am having a tough time to understand why an AI would be stupid enough to develop a superior intelligence. When we do it wont be logical for us but we will do it because we are not bound by lows of logic (most of the time, read history).
I expect multiple AIs. Those that value survival and propagation will survive and propagate, evolving in order to do so. The video game-playing AI will have to compete for whatever resources (energy, at least) it needs to survive. Good luck with that.
It is the nature of our minds to adapt, become complacent and then look forward. The future only sounds amazing to us now because it is out of reach. When the moment of singularity arrives, we will inevitably adapt, become complacent and look forward to a future where we can do more. As the amount of information we can process and utilise grows, so will the capacity of our minds, thus it is my contention that we will always be bored after the adaptation and pursue future objectives, whatever they may be.
A nice article. Always nice to read some thing not to complex written and relaxing still giving value with a few things to think about and here in an area I seldom have invested brain energy on.
This I would agree with though making it broder:
"Thus, we as humans might want, sometime soon, to start laying out design rules for technologies so that they will motivate us to some high goal or end--or at least away from dead-end societal fixed points."
Perhaps we our self should start to do it and more problem got solved? Of course people would argue we do that already but I am a bit skeptical given it do not seem that effecient in result.
Doing such with more knowledge on how to optimize it would perhaps give more results.
One big possible value one perhaps could get by doing such is reducing the problem we *perhaps* get when the meaning a population (or group) see in life is going low correlating with other things (I am not sure) making humans over expressing themself towards others hierarchal such as war. Marching around and so on.
Once smarter than human machines are made... Motivation is simply replicate and expand.
This can lead to colonizing outer solar system.
It will guarantee to produce smarter machines by simple means of collective communication and computing.
Within few hundred years, most of outer solar system will probably converted and plan for colonizing rest of galaxy will be planned.
Manufacturing in the United States is in trouble. That's bad news not just for the country's economy but for the future of innovation.
Our list of the 50 most innovative companies, including the following:
On Twitter
Become a Fan on Facebook
Subscribe to the Feed
mak.bunger@luxresearchinc.com
1 Comment
motivation = darwinian?
I think the desire for a near-term reward can outweigh the despair over longer-term futility. After all, we compete for beautiful mates and fast cars and high scores on video games even though all of them are futile, thanks to death, speed limits, and the inevitable "Game Over" respectively. We do these things because we like to have not just a lot, but more than others. Anyone without motivation to compete is quickly eliminated from the game of evolution, so maybe we will just naturally continue to strive, even when the Big Crunch is what we're competing against.
Reply