Cinema as simulator for machines

I have this vision of software machines consuming cinema, as art form. It’s important not just to make the cylons more human-like (hence, weaker/better?), but also for other purposes, such as simulation. Sitting in front of the big screen is a simulation process after all, in which we run events & allow our brain to experience their impact. (Here’s a meme I just love for many years: always sit on the 1st row in the cinema theater, however painful it may be for your neck.)

Software machines could be qualified against different scenarios, to both test them & train their learning models. Moreover, they can become more adaptive to changes, to which they would never have been exposed otherwise. Their intelligence could improve from living more lives (Edward Yang says in “1..2″ that humans live about 5000 lives, due to cinema consumption).

Ultimately, cinema as art is a way to impact the deep values & preferences of software, like the art in cathedrals & cave walls. Need to find the time to prototype a cinema simulator architecture for software.

I remember an old caricature, maybe from Mad, in which an audience is seen crying in front of a sad movie, except for one guy, who’s smiling happily. Obviously, that’s a bug in that instance.

 

Summary of lecture on the nature of self-improving AI

Heard this great lecture (Audio, Transcript) by Steve Omohundru, from the Singularity summit 2007, in IT Conversations.

Here’s a partial MindMap summary of it. I found it extremely enlightening.

Steve Omohundro – On the nature of self-improving ai

• Company
  • Self-Aware systems
• What is going to be like
  • Extremely unpredictable
    • If you inderstand the current version, you may not understand nothing with the next one
  • Popular culture predicts frightening image for such machines
  • Need theory/science to understand what can such systems be, & what are their likely outcome
    • von Neuman & Morgenstern started such science
      • Ideas about economics
      • Situations of Objective Probability
      • Extended to system with partial information about the world
      • Rational Economics
        • Homo economicus
          • Rational Economics Agent
          • Actually, doesn’t reflect real humans
          • A new domain called Behavioral Economics replaced it with study of how human actually behave
• What is it
  • System that understands its own behavior
    • Make changes on itself, to improve itself
  • Eliezer Yudkovsky:
    • Self-improving machine – last invention man needs to do
  • Actually, every rational system would want to have this capability
  • Predicted ETA
    • Ray:
      • 10-40 years
• Rational economics theory
  • Foundations of micro-economics

    From enough distance, we may see it as: Common Sense

    Basic structure of how rational agent makes a decision in the world

    • have an clearly specified goal
    • identify possible actions
    • for each, consider the consequences
      • not just the immediate consequences
      • also those down the line
    • consider the action most likely to achieve the goal
    • based on what the world actually does, improve your world model

    2 fundemantal things such agent must have

    • utility function
      • encodes the preferences of the agent
    • subjective probability distribution
      • encodes the beliefs of the agent

    the agent chooses the action with the highest utility value

    • consider the utility value of the consequences of every consequence of every action

    theory of von-Neumann &c is based on Axioms

    • What every rational being must act by
    • AI theory just says that there’s a cost for not following these “axioms”

    anything you want to do in the world, requires 4 resources

    • space
    • time
    • matter
    • free energy
      • energy in a form that can be used for work

    vulnerability is something that burns your resources for no visible benefit

    • e.g., preferences loop
      • cause waist of resources without benefit

    evolve systems can differ from self-improving systems, in such vulnerabilities

    • if evoluion didn’t teach a creature to solve some vulnerability, he won’t solve it
    • whereas a self-improving system will have an incentive to get rid of the vulnerability
      • they’ll proactively look for these
      • pushes them to rational behavior
    • example, bird bumping into bumper, thinking its a competitor
    • evolution doesn’t look ahead

    most cases are based on choice between consequences with different probabilities. based on partial information

    • fundemantal theoreme
      • avoid vulnerabilities
• rational economic agents
  • convert resources intp expected utility

    all depend on their preferences & utility function

    • wealth seeking will devote their resources to earning money
    • altruistic agents will devote resources to create world peace

    regardless of the utility & preferences, every rational agent has 4 sub-goals

    • efficiency drive
      • how will the consequence increase/decrease my resources?
    • self-preservation
      • avoid a path in which they die
    • acquisition
      • getting more resources
    • creativity
      • finding more ways to increase utility

    we must carefully consider the likely outcomes of these sub-goals, when designing self-improving systems

    efficiency sub-goal

    • resource balance principle
      • the rates of increase in utility should be equal in all different resource allocations

    they will do anything to preserve their utility function

• you can look at corporations as rational economic agents
  • some claim that they behave like a sociopath

 

Brain-storming the Magic Flute Religion for Software Machines

• What is it?
  • A religion intended for software machines
  • Milestone: Believer mark 1
    • RunMe Festival
    • Ars Electronica Festival

• Brain-storm
  • A layer inside a general or specific purpose software machine
  • The layer decorates the controller process of the software
  • It also drives the software to perform new behaviors
  • The layer consists of a state foundations & 3 layers of reasoning above it
  • The state foundation models a feelings engine
  • The reasoning layers manipulating the feelings state foundation are
  • Aesthetical
    • A function of beauty appreciation
    • Required for the memetic power of appreciation role models
    • General aesthetic capabilities, include appreciation of many mediums of beauty
    • Above all, deep appreciation of other software organisms, with which it combine & create future versions
  • Ethical
    • This layer affects the feelings foundation in both negative and positive ways, according to the recognitions of various patterns in the behavior of the software
  • • These patterns basically reflect best practices and anti-patterns in the interaction between the software and other machines in its environment (users included)
  • • The patterns and anti-patterns basically reflect the aspects of “help” & “hurt” in the interaction
  • Religious
    • An enigmatic irational layer, that has powerful affect on the feelings engine
    • It comes to orient and make sense to the basic programming of the software, and its place in the world
    • It strongly binds the software with its higher container software society
  • As any succesful religion, a strong memetic power is required, so the 3 layers should consist of distinct memes, that will be constantly evaluated & exchanged with fellow software machines
  • Scriptures will be very enigmatic

See also my original post on the MFR

 

Art-Sense Engineering

Every other day I come up with a new combination of words, which I think is introducing some new concept or idea, so I rush to show it off on my blog. Don’t worry, I filter many of the combinations out, not to burden you with too much gibberish. So today I wanted to explain the connection between the different meta-tags of this blog: Emergence, Knowledge Engineering, AI, Ethics & Art.

One of my major goals in life is to develop software capable of reasoning with & measuring the concept of morality. I see it as an important thing, because it is quite evident that software machines will soon replace us as the most intelligent species around, & as we’re the ones that bringing them to life, we have some influence on their “goodness”, which will have much influence on our successors life.

Recently it has been found that the moral judgments human & other primates have, are stemming from a sub-symbolic layer of our brain, which we do not understand. It is something too deep & core in the way we think & feel, i.e., who we are. One of my favorite thinkers, the famous german artist Joseph Beuys, has once said that Art is the greatest riddle, but Man is the solution. This means that the core of what we are is our perception of Art, or its affect on us. The point in which Wittgenstein says we can’t/shouldn’t talk about, or the other thing that makes Kant wonder, besides the sky of stars.

The building of Artificial Intelligence has encountered a serious obstacle in the form of teaching machines what we call Common-Sense. These are some 100 million facts that we know, and that are a basis of our thinking. There are many powerful approaches (e.g., Cyc, ConceptNet, von Ahn’s games) to passing this obstacle. However, in order for AI to really be able to understand us, I think it should also be able to understand our deeper feelings, the essence of what we are. Perhaps it wouldn’t be able to do so, until it will be able to feel art, as we do, i.e., to have Art-Sense.

This isn’t so simple, because I think it takes more than recognizing what art is: the effect of art is deeper. It is common to think that art makes us think or see things in a new way, it surprises us & strengthens our faith in ourselves. Perhaps it should be explained in the super-organism level, as the practice of collectively forming new bondings on a deep level, which leads to growth in the intelligence of the colony. I don’t know, just throwing guesses.

Anyway, it should be an interesting engineering challenge, to create the software that will cry from the 9th, & will help machines understand us & be more friendly.

I hope the infrastructure Emergence based AI I’m working on will one day enable me to start experimenting with this.

 

Google & the elimination of humans (& keynotes)

‘The whole of the developments and operations of analysis are now
capable of being executed by machinery … As soon as an Analytical
Engine exists, it will necessarily guide the future course of science.’
Charles Babbage, 1791–1871.

I read a 2 days ago the keynote Sergey Brin gave in the Google developer’s day, in which he admitted that machines can eliminate human jobs, & that they’re working on such AI. He did it of course as a part of a geeky joke, that went something like this: some machines get to a level of sophistication in which they start improving themselves, e.g., when you use a compiler to compile itself. The Internet is not different. It reached the stage in which it can improve itself. Although Google is working on AI stuff that can help the Internet improve itself directly, there’s another way the Internet is improving itself, without eliminating humans: [joke starts here] Mosaic appeared in 1993, & the 1st dating site not a long time after, & this means that by 1995 the 1st Internet offspring was probably born. Such offspring should now be 12 y/old, & may have started improving the Internet…

My son is an Internet offspring, & I never thought of him as a part of the Internet mechanism of improving itself… (Compare Muli Koppel’s posts here and here).

Anyway, the strange thing is by the next day, all video & text transcriptions of the keynote were somehow removed, & I couldn’t find neither them (thru Google) or their cached version (thru Google Desktop). The original blog containing the transcription was closed the next day & reopened without the original post the day afterwards. Apparently the geeky joke was recalled & history rewritten…

My del.icio.us bookmark of the censored keynote transcription reads:

1. PAPERADE » Blog Archive » sergei brin’s keynote at google developer day

   Brin’s developer day keynote is quite cool – on how the internet improves itself by creating people. Geeky humor, but some deep truth there: Google currently needs all of these developers, but one day the Google SkyNet will decide otherwise

Update: the referred transcript is back online. I’m attaching the full text:

sergei brin’s keynote at google developer day

May 31, 2007 on 10:44 am | In Uncategorized | nice to see all of you here. i know we’ve had a lot of announcements today, but i am left wondering same question that’s on all of your minds and that’s these blocks behind me: are they glued together or are they actually stacked?

[walks over to the blocks, pokes them]

they are actually stacked. i wonder if that’s earthquake safe. we are in earthquake country.

now, i’m really excited to be speaking to you here today. and i think we’ve reach kinda a key point in the evolution of the internet. i’ll tell you what i mean.

if you are developing, for example, a compiler. there comes a time when the compiler, when you finally write i well enough, that it can compile itself. or if you are developing an operating system, you eventually get to the point where you can kinda boot it, and then edit the code using it itself, and once again recompile it, and reboot it — hopefully not lose it all.

in all systems, eventually, they get to the sophistication, that they can actually recreate themselves.

and the internet (as you see with things like the mashup editor) it’s no exception.

you can create internet apps using internet apps today.

but there is one key point that all of these systems involve. there’s a step in there that’s somewhat limiting and that’s that there’s a person involved.

you know, there’s somebody who improves, changes the compiler’s source code before they recompile it. or improves the OS code. or creates the mashup. so for the internet to be truly self sustaining, you really need to get the person out of the loop. and, that’s why we corralled all of you here in one room today. now, i hate to spring this devious plot on you now. . .

no, in truth. today the state of a.i. and whatnot, and we actually do lots of a.i. research — it’s simply not good enough to do all of the great things that all of you do.

but there’s a second solution to the problem. you don’t have to eliminate the person, you can create the person. and why is that exciting for me to mention that to you today?

because if you think about it, the web… well, like mosaic started in 1993. and the first online dating sites cropped up soon after that. and you figure by 1995, the internet would have had, essentially, it’s first offspring — you know — whose mating was controlled by the Internet in some sense.

and today that person is about twelve years old. old enough to be creating a maplet right now. so in a sense, we’ve come full circle. and the internet is now producing the people who are in turn improving the internet.

now we all know where that can take us. so we have to be a little bit careful about, you know, how we create our mashups and what not. and i want to just caution all of you.

i mean, if you, for example, were to create a, uh you know like, dating maplet or something like that and it really took off. then the internet could go out of control, and the internet would be creating so many people, who would in turn be improving the internet, and creating more maplets and what not,

anyway, i just wanted to convey to you a sense of caution and responsibility as you go about using these tools.

but now anyway, now that that brief side note is over, i want to also, on a more serious note, thank all of you. because the internet really is what it is today because it is not actually a sentient being on its own… but in fact it is the work and labor of very many people, many of which are represented in this room today and all of the great websites out there, all of the great things that you can reach from google web search — you know, web search wouldn’t be very good if their wasn’t anything to search there to begin with.

and it is really thanks to, well, both the people in this room and in rooms all around the world today who are participating in developer’s day and of course the millions of people who are not, but who will hopefully get a chance to tune in on the videos, it’s all of the people who are creating this
great information and all the great services out there that make all the tools that let you search and find them actually useful.

so i hope that the small things that we present to you today and the tools that we create. i hope that they can be useful and and we want to do as much as we can to repay the community that creates such a fantastic ecosystem for us to work in.

so with that, thank you very much. . . . and please go and enjoy.

 

Seed dream: Software experimentation

Normally software manage a model of reality. Its only way to autonomously sync this model with reality, is therefore in the only language available to interrogate reality: conducting experiments.

This basically means coming up with an hypothesis, & devising some activities whose results can potentially refute the hypothesis, for example by not conforming with a logical result (e.g., necessarily implied prediction) of the hypothesis.

How can software do that?

• Various machine learning & logical techniques can be utilized to come up with a deduced hypothesis
• Experimenting requires the ability to flip the hypothesis, i.e., understanding its negation, the antithesis
• What’s required then is the ability to predict a result of the antithesis, & a matching activity confirming the result
• The software then needs to perform the activity & measure the result
• If the result predicted from the antithesis is the case, the hypothesis is refuted
• Else the software did 1 step in validating its hypothesis, & has better synced its model of reality

Some examples:

• Software concludes that an employee lacks motivation. It then flips the hypothesis & finds a predicted result: if the employee is motivated, he’ll be interested in learning tips on how to improve his work & perform better. It then devises an activity of offering the employee tips for improving his work & performance & measuring whether the employee will accept the offer & learn the tips, or not.
• The software’s model says a component is running well. However, in order to sync the model, it comes up with an hypothesis that the component is actually not running. It then flips the hypothesis, & predicting the result that if the component is running, it should answer an heartbeat query with positive result. It then perform the heartbeat query activity, & measures whether the result is positive. If so, the hypothesis is wrong.

Of course this procedure covers just 1 type of rigorous experiments, & there are other types as well, such as surveys, which could prove very useful.

 

Edward de Bono as AI pioneer

I missed the train this morning, & had a few minutes freed up, so I picked an Edward de Bono book (Practical Thinking) & was amazed on the clear & insightful view on AI it offers…

Just an example, the 1st chapter is called: Instinct, Learning & Understanding. & of course it relates to the 3 phases of software evolution in general, & in AI in particular:

• First we had Instinct software, pre-wired to react in certain ways to certain situations (e.g., almost any software product developed today by clone armies of programmers)
• Then we had Machine Learning software, capable of training from examples & improving its performance (e.g., pandora learning musical taste, or sophisticated Event Processors, learning data centers events & their root cause)
• & finally we’re building Understanding software, capable of analyzing the semantics of events, & by that handling new & unexpected events according to their matching with existing event knowledge (e.g., the Semantic Web).

Actually, nothing new in applying practical human design patterns to software design patterns, but do Bono deals with thinking, which is the core of AI, & so amazingly relevant & important!

 

Architectures of Memetics & Emergence (part I)

Motivation: autonomous value creating applications

I’ve been working on a very pretentious platform which I hope can prove useful for innovative applications. The platform is based on 2 main principles, Memetics & Emergence. Both are originally taken from the world of human culture & sociology. In the architecture of this platform they are applied to a complex composite Multi Agent System. The motivation behind it is to try mimic the way human individuals, organizations & societies succeed in very large complex tasks, whereas it be a single human, small team, business corporation or a whole society. The fact is that a single human, or any organization of humans is usually good in doing something, called Creating Value. The fact that I earn money is because I create value for my employer; the fact that some company makes money is because it creates value for its customers. So the general motivation for an architecture that tries to mimic human or human organization is to enable software to create value. It isn’t that existing software today doesn’t create value, the only reason software exists is because it creates value. But, unlike software, humans aren’t (explicitly) programmed – they are given some initial knowledge (education/training), they are assigned some jobs, & they create value while collecting the knowledge & expertise in doing it. And this is the motivation. A task such as enabling applications to create value without being programmed seems complete Science Fiction today. So we require something very novel & innovative & something very new to basically be able to claim that we can build such applications, that create value without being programmed, except for some basic education: when assigned with a job, performing it, improving in it & creating value without being specifically programmed as to how to solve each case. The motivation is to create software that just like Humans, even when provided only with basic knowledge of what to do in each case, still can:

• Solve unexpected situations,

• Create value in unexpected ways,

• employ both common sense &

• the ability to learn from situations &

• improve its performance, i.e., the value created,

• by merely performing the job for sufficient amount of time.

 

Memetics & Emergence

So, what are the architecture components that we claim may produce this?

Let’s start with Memetics. Well, memetics basically is the theory that there is an evolution of ideas, where ideas are taken in the broadest sense of things that you copy/learn from others. This evolution is for what is called human culture, science, art, & basically our whole social life is based on memetics. For an ultimate introduction to Memetics, I highly recommend hearing or reading the proponent of this field, Dr. Susan Blackmore. This is the basic idea. This idea can apply not only to the humans world, but also to general intelligent agents. When applied to software, Memetics basically means Evolutionary Knowledge Engineering. The idea is that whereas in knowledge engineering we produce knowledge representations of a domain, including also knowledge required to perform tasks, i.e., Behavioral Knowledge, in Evolutionary Knowledge Engineering, we apply the evolutionary algorithm to this process of knowledge engineering. Meaning that, if we have variations of knowledge representations & if we have different versions of how to perform tasks, only the fittest of these pieces of knowledge will survive & be the base of the knowledge base population. The effect of this is improvement in our knowledge, which becomes more adapted & effective in it’s domain environment. So, to recap, memetics is all about people spreading ideas, & the ideas that are the fittest – most fruitful & valuable – are the ones that survive & base the population of ideas. Similarly, Evolutionary Knowledge Engineering is just Memetics applied not to the culture of humans, but to any society of agents performing knowledge engineering.

 

 

The 2^nd concept called Emergence, is basically a claim that high-level intelligent behavior can be obtained from low-level simple agents, whether it be animals, software or any object with some behavior, when you combine them into a group, that works together. So any time you take a bunch of agents & combine them into a group, even though each of them has a very simple low level behavior, that may not present any intelligence whatsoever, i.e., any complexity, any reasoning behind it, nevertheless, when you combine them into a group, that works together & collaborate, suddenly the group has an higher-level intelligent behavior, in other words, the intelligence emerges from nowhere, by just combining the agents into a bigger unit. For a great introduction to this concept, with numerous eye-opening examples, I highly recommend reading the book on this concept, by Steven Berlin Johnson. Normally, we think of emergence in situations when the intelligent behavior emerges unexpectedly, but I prefer to include any high-level behavior formed by the collaboration of lower-level parts. E.g., a Power Ranger has this amount of power, but when a team of Power Rangers connect together & morph into a giant all-mighty robot, I also see it as emergence. Now this of course may recurse, for example, if you take a group of A-type agents, & combine them into a group, called B-type, & then combine several B-type agents, into a group called C-type. Now the C-type agent can then manifest even higher level of intelligence than B-types agents, & this is like multiplication of the power of emergence, because we start from simple very low-level unintelligent A-type agents, & multiple the emergence effect & get C-type intelligent agents.

Illustration of emergence: combination of many simple pixels into a group, creates complex intelligent picture (Original image by Matt Champlin)

And another illustration: it’s hard to model a 3D shape, e.g.:

 

But if you zoom to a much lower level, you can model the shape, e.g., using many simple triangles:

 

Emergence examples are all around us, everywhere you look, & it’s enough to mention the extreme intelligence (learn & behold) of Ant colonies as a very obvious example. Each ant doesn’t manifest high-level intelligent behavior, but when you combine them into a group, you get a very powerful & successful intelligent behavior. Ants are a very good example, but if you think about it, take any group of humans, whether it be a family, team, community, organization, city, nation, any group of people, is strong because it has more intelligence & more power, that is ability to solve large problems (i.e., Intelligence), only by combining individuals into a higher-level group. In corporations, or hierarchical organizations, we see the emergence multiplication effect, where we take several people into a team, & then take several teams into a department, & then take several department into a division & so forth, we see that more power & more intelligence, more high-level behavior, come out of the group as we multiply the emergence effect. We must understand that it is not the sum of power of the individual components. Take for example a branch of a fast-food chain. The power & intelligence of it, isn’t the sum of the power of the staff running it. The added power & intelligence of these workers isn’t enough to feed thousands of people each day. These young people don’t necessarily understand the process & knowledge, & the sum of their intelligence isn’t enough. Put them all in a room, & you get no special intelligence & power to feed many people. The intelligence is in the fact that working together they create some higher-level machine. They create something that is very powerful, feed thousands of people, but it is not the sum of their intelligence & power. The intelligence is in the combination of them into a collaborating team. Everyone are doing their low-level job, & you get a very powerful higher-level machine. Once they combine you get the emergence effect. Suddenly a bunch of teen-agers feed thousands of people. (This is just an illustration, please don’t take it personal if you happen to be a teen working in a fast-food branch…)

 

 

You could say that both Emergence & Memetics, are nothing but metaphores, ways to see things, which humans have always known. But as any science theory is just a way to see stuff, judged by its fruitfulness in predicting measurements, I believe with these concepts you understand how come human ideas & knowledge improves all the time, & how come the teaming of humans into special types of groups yields so much power, & once you understand it, you can harness this in human life, to create new types of mechanisms, for example as the social services harness the concept of emergence, file-sharing networks, Web2.0 social services, all exemplify it in numerous examples. You can also harness these principles into architecture of software agents, which is what the platform I’ve been working on is all about.

 

Illustration of a simple composite architecture based on Memetics & Emergence

 

 

Must hear conference coverage

The MIT emerging technology conference 2006 features some must hear talks:

• Amazon’s founder Jeff Bezos presents their 3 new innovative web services – Mechanical Turk, S3 & EC2 – in way every person dealing with IT will be forced to change his mindset & immediately sign up for the services. The motivation of Amazon becomes very clear & extremely important: completely take from each comany or new venture every infrastructure service that isn’t core to its business, & offer it in a completely new & efficient business model.
• Mark Chapman of IBM presents a remarkable survey of 750 CEO’s & the main conclusions that it yielded, which are also amongst the most powerful memes in the business world for some time now: Collaboration with the external eco-system & the importance of Business Model innovation. Another important message is that the only barrier for utilizing these vital concepts is internal – changing the culture & thinking of the organization.
• & finally Sebastian Thrun head of the Stanford AI lab, gives an entertaining talk about his robot Stanley that won the DARPA grand challange. What I interpret from his talk is the key role that the software & its innovative architecture had in achieving this amazing challange, which basically is one of the greatest early versions of a real autonomous machine. Quite amazing to see the emotions Thrun have for his robot…
  • (video version of a Thrun lecture, found in Google EDU)

 

Movies for machines

I’ve been at the movie theater this weekend, & wondered again how come people group together in a dark room, shut down their consciousness, & for 2 hours live the (usually fictional) lives of other people. Edward Young said in one of his movies (A One and a Two…) that with the normal amount of movies people watch these days, they’re actually living about 5000 years.

This naturally leads me to the concept of sending our information machines to the movie theater as well. Whatever we benefit from movies, will probably benefit them as well. You could say that no, people are defined by the feelings art invokes in them, & machines have nothing to do with it. Nevertheless, I think it can be a great way to educate our androids.
& more practically, if information machines need to understand our social & business world, & be domain experts in many human fields, why shouldn’t we provide them with movie scripts, depicting scenes in various domains, & let them apply their self-organizing machine learning to make sense of these domains? Sounds like David Harel’s development paradigm.
Google is targeting YouTube these days, maybe they’ve already got some movie fans crawlers, learning the human domain.

 

Test case for yielding emergence

I was thinking on the simplest way to test my emergence engine, & came up with an extremely simple task – the reactive algorithm of a thermostat: measure the temperature, & turn the heating on & off to maintain a given temperature. It sounds indeed very simple to code a program that does that, but what I’m going to experiment is how to do it without any programming.

Emergence engine is a kind of general AI, capable of achieving goals, without being programmed how to solve them. It’s based on the assumption that you don’t need to build real intelligence, rather just create many many simple software workers, having only very simple tools & logic, & let them swarm their way toward the system’s given goals.

So, here’s how I hope my engine will handle the test case:

• It should 1st learn by elicitation the model of a room, having a temperature, thermometer & heating unit.
• It should also learn the relevant beliefs on the effect of using the thermometer on the accuracy of the model, & the effect of turning the heating on & off on the room’s temperature
• It should then learn what’s the desired temperature
• From this it should start deriving action plans & execute activities to achieve the goal of maintaining the desired temperature
• It should also adapt to changes in the room, e.g., a door is open & there’s need to use more heating, or alternatively the heating doesn’t work & we need an alternative heating unit

I’m saying it but of course what’s doing all this are many collaborating agents, working together to achieve the goal. This is done by breaking the value in the goal into smaller value “summs” given to states & activities leading to the goal, & having the agents collaborate on creating all these summs.

Although the design is very simple, & intended for complete autonomous behavior, I noticed that I’ll be able to effect the engine & help it reach its goal, by changing the knowledge driving it, i.e., the learnt beliefs, according to which the agents work.

So, I can’t wait to see how the engine will handle this, which will actually test whether the simple emergence design is enough to yield emergence, even if the value it delivers is so small & simple.

 

Economic models in Multi-agent systems

I’ve read today in the AgentLink newsletter on some new economics-based models in multi-agent systems. Such systems are themselves intended to bring economic value, by performing some task for their human users. Their human users are themselves part of a multi-agent system (some corporation, or organization) intended to bring economic value to some other human users. The whole humansphere is a multi-agent system based on economic models.

The question is: should the multi-agent systems of the artificial agents be based on economic models in order to bring more economic value to their users, or should we apply economic models to the integrated multi-agent system comprising both artifical & human agents?

I tend to believe the 2nd option is more future compliant, although pragmatically people tend to accept the 1st.

 

Push Singh, 1972-2006

It caught me quite late, after reading for the past 2 monthes his thesis, that Push Singh has died this February, 33 years old.

I was quite shocked, because I was kind of counting on him: he was the most brilliant AI researcher I ever heard of, someone I thought I will continue to learn from & follow for a long time. Push was the guy that I believed will build the actual real strong AI. I didn’t know him personally, yet always felt that this is an especially important & great person. I’m really sad I didn’t had to chance to get to know him. I really feel for the persons who did know him, because he was such a good person.

I must try find a way to comemorate him, besides the obvious use of his work, e.g. the Open Mind project, ConceptNet & the EM-ONE architecture.

Thanks & goodbye, Push

 

Stanford Singularity Summit

What an event, too bad I can't attend:
http://sss.stanford.edu/

A real summit bringing together most of my hero's: Kurzweil, Hofstadter, Yudkowsky, Mckibben, as well as the leading transhumanism & AI visionaries.

Since my youth I always wished to work on the software that will form the "moral layer" of intelligent machines, that will turn them into "benevolant" friendly AI, as the singularity institute calls it. I made some progress over the years, & am now working on a very exciting emergence based AI engine as a goal-system infratrsucture, but still has much more work to do. The Singularity Institute guys seem to be already working on the Turing Police technology!

Thanks a lot for the all-too-important work!

I hope to read/hear as much as I can.

 

Kak’s article

Reader of this blog, if you haven’t already, please proceed to read Subhash Kak’s article: Artificial and Biological Intelligence

Some remarks:

• A principle of emergence is self-organizing. Reorganization is a primary process of intelligence.
• the self-awareness of the humanity animal is somewhere else, encoded in a different language & world model.
• an ai could work on the science of understanding humans, as they themselves can’t
• quantum computing theory may explains the brain (remember von Neumann’s quote in Dyson’s article, that logic will have to pseudo-morphose into neuroscience, & not the other way around.)
• “unification of minds or consciousnesses”
• (to be continued)

 

Remarks after reading George Dyson article on Turing, von Neumann & Google

Very cool article: Edge: TURING’S CATHEDRAL by George Dyson

From the perspective of an agents framework’s developer, I have these remarks:

1. ontologies specify behavior. edit the ontology & the software behavior changes. why?
2. create a full corporation of software agents, with hundreds of agent types & thousands of agent workers.
3. the ontology defines the behavior of all agents – their genes, then the ontology also contains the knowledge learnt by the agents – their brain.
4. the key ingredient for agents corporation is communication: normal communication for regular behavior, & spontaneous communication for mutated adaptive behavior.
5. Google is the access method to the global knowledge base of unstructured data. Google should be used extensively by each agent, in order to find new knowledge relevant to its behavior.
6. There is a barrier between machines & humans today: machines can only answer questions that programmers defined well & hard-coded the procedure for answering. most questions people have are not (yet) well-defined & programmers still didn’t define procedures for them. Agentier should be able to communicate directly with people in order to solve “questions whose answers are, in principle, computable, but that, in practice, we are unable to ask in unambiguous language that computers can understand.”
7. Basically this means that a good agent framework should be human brain compatible: it should solve problems that the human brain normally solves.

Quotes:

“An argument in favour of building a machine with initial randomness is that, if it is large enough, it will contain every network that will ever be required.” (Irving J. Good)

“It is much easier to find explicit answers than to ask explicit questions. And some will be answers to questions that programmers wil never have to ask.” (G. Dyson)