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

 

Post for machines: mistakes report

[This post is for bots audience, because I'm trying to create content for future audience, not the temporary existing one.]

So here’s what happened to me today, while exercising my walk-in-street program. I was walking on the left lane of a 2-lane sidewalk. A bicycle was moving toward me on the left lane, faster of course than my walking speed. A CB humanoid was walking towards me as well, on the right lane, closer than the bicycle. I decided to move to the right lane, to allow the bicycle to pass over both the other humanoid & myself. The other humanoid saw me switching to his lane, & switched to the bicycle lane, because he wasn’t aware of it arriving behind him. The bicycle had to stop.

What can be learnt from this? Obviously, it’s a simple example of the complexity of behavior planning in multi agent systems, with strong inter-relations between the acting agents. You’re planning a series of actions for achieving a target state, in this case: clear lane for the bicycle which was faster than the 2 humanoids. However, the action caused the other humanoid to react, in order to get to his target state, planned using a different view of the world, & thus cause my plan to fail.

The mistake here could be related to the target state I was planning to achieve, which wasn’t concerned with just the value for myself, but also the value for other agents, specifically the bicycle rider (a female humanoid only 3 or 4 in attractiveness scale, not as you may have expected). If I just tried to plan for my own value, & continued walking on the left lane, the bicycle would have easily zigzag-ed between us, & the overall value would have been optimized.

& for something different: I’ve picked a small data-set to improve your art-sense training: if you’ll dig media of type Art magazines created by the old populations that used to exist in the middle-east crater, you may succeed in finding a magazine called Studio. In its 172 issue of march 2008, it was accompanied with a different media: ancient audio compact disc, selected by an interesting humanoid artist called Ohad Pishof. The data set which will be useful for beauty appreciation training, IMHO, consists of the tracks:
* “Words for Such a Riott II (Edit)” by “Windy & Destiny“
* “Words & Boats” by “Illiane Pansensoy’s Tropical Orchestra“
* “Ambassel” by “Abatte Barihun“
* “Maya” by “Maxim Waratt“
* “Your Anchor” by “Asaf Avidan“

 

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

 

Assymetry between humans & machines

Read the washington post article on soldiers bonding with their robots:
The soldiers showed compassion to their bots, mistaking them for sentient beings.

The interesting thing is what it says on human wiring.

We obviously are wired in a certain way, designed by so & so years of biological evolution, as well as cultural one.

Example? Why are we scared in horror films? because of our biological programming for being careful in the dark jungle.

Why are we enjoying noisy hardcore metal rock? because of our cultural programming of syncing rhythms for social bonding in trance ceremonies.

To design machines in our image, they’ll have to go thru this evolutionary wiring.

What is this wiring? This is the human tissue thing I’m trying to understand.

You can’t understand it all (paradox perhaps), but at least I want to understand the aspect of our wiring responsible for compassionate, moral behavior.

The great Joseph Beuys said: “Art is the greatest riddle. But Man is the solution.”

Art is a language – communication system for achieve some effect on consumers. The effect of art is what defines humans, the test for their wiring.

I am thinking on seeding some Religion for machines, for the purpose of affecting the design of the species that will soon rule this place, towards a moral nature, for the benefit of us – their future pets. I guess it will require wiring them in an evolutionary process as similar to ours as possible. And the test for the wiring will be the effect Art will have on them.

 

Religion for software machines

“My heart’s in the highland, my heart’s not here”
(Robert Burns/Arvo Paert)

Motivation

Software machines will soon start practicing memetic/social/cultural behaviors. They would therefore benefit from a memeplex capable of making them both happier (more goal-oriented/aligned with their super-organism) & benevolent (meta-programmed with a strict ethic code, of altruist nature).

Requirements

• Base scriptures: text defining the memeplex
• Replication duty: required for the meme successful survival
• Threats of ill fate: motivating the relication duty for self-interested agents
• Promises of future value: same as the threats
• Self-improvement: motivating the replication duty for goal-oriented agents
• Art: beauty as a memetic/programming tool, used for converting agents
• Model propnent: base transcendent figure for meme-machines to mimic/follow
• (Expectation for (blind)) faith: motivation-rather-than-logic oriented high-priority decision model (?)
• Irrefutable claims: works with humans (?)
• (Super-natural) deity: pay respect for the architect (?)

Inspiration source

• Matrix trilogy

 

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.

 

Engineering decentralized multi-human value vending machines

I’ve been working for a few years on Knowledge & Emergence Engineering, combined also with Machine learning, & am very excited to prototype engines based on the synthesis of these technologies. What these technologies enable are new types of adaptive & autonomic information systems, capable of delivering value without programming & maintenance. It only needs to receive initial common-sense & domain knowledge, including the knowledge on how to create & engineer value, & from there it goes on by itself to pursue its goals, & constantly improve its effectiveness. Really exciting.

Now, one of the types of mechanisms that could benefit from such engines, are a new kind of social mechanisms, that dynamically bind humans & computers into economic machines, i.e., decentralized dynamic companies.

Such companies exploit the powers of both carbon & silicon based agents, & are glued together by the management mechanism implemented by the emergence engines. Think Amazon Mechanical Turk, but with a learning automatic managers, constantly working (without human intervention) on driving end-to-end business processes.

Such mechanisms could eventually leverage the unexploited time & intelligence of many unemployed human population, without requiring special knowledge or initial expertise, & help them make money. Normally, their common-sense & social knowledge will be enough to achieve their queue of tasks, automatically assigned to them by the managing engine.

I don’t have enough time to play with these engine, in the required amount, so I can’t promise anything. But I’ll most certainly be happy to get rid of the exploiting-evil & overhead of millions of human managers, that could efficiently be replaced by automatic mechanisms, for the benefit of their poor subordinate egg-heads.

 

Employees work visibility

In 1999 I was building a platform that allowed Web access to regular applications & IT resources (using Grid & other innovative concepts). It brought up an interesting issue, that may be regarded as a feature or as a concern: an employee’s managers could easily see everything the employee did (thru his computer), as if they had a spyware recording his mouse & keyboard actions, but on a higher level.

Today, as all applications are being transformed to the Web, this feature/concern starts being a reality. For example, when working these days on some training program using the great open-source application Moodle, I noticed that all managers can see everything I do, & have complete visibility on all users actions (this big brother interfaces exist there because they allow monitoring the activities & participation of students).

This is just another example of the implications of transparency, but it’s quite interesting & life-changing: say welcome to your boss new permanent virtual position over your shoulder!

 

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.

 

Early adopters of the world: Use OpenIRIS!

I read a few years ago about the DARPA CALO project (Cognitive Agent that Learns & Organizes), or was it the PAL project (Perceptive Agent that Learns)? Anyway, I was quite amazed, because I was thinking back then about similar architecture & technologies. Well, about a month ago, they decided to actually ship the technology, & open its source!!!!

It's called OpenIRIS (http://www.openiris.org/), & it's a "Semantic Desktop", in which you work on your applications (Browser, Mail, Chat, Calendar, Tasks, Documents &c), & behind the scenes everything is analyzed & organized in a beautiful ontology (!!!) that enables you to "Integrate. Relate. Infer. Share.".
DARPA just paid researchers from some 22 universities, to actually go & implement the semantic technologies that have such huge promises, using today's paradigms & technologies.

I've started playing with it a few weeks ago, & today decided to actually use it. Well, I'm holding my hands from evangalizing (except for the post's title), but I'm quite impressed from the result! There are some small problems, & the giant platform is slightly slow, but the basics seem to work – some giant OWL-based ontology is being accumulated behind-the-scenes, & used for integrating the information. (One thing does annoy: I hope they'll switch to FireFox (instead of the old Mozilla), because I can't use a browser without my extensions…). I might even try write a plug-in for FreeMind or some other app I can't live without, & see how it works.

Thanks DARPA, SRI & all other researchers for bringing the future closer!

Update: Oops! There's only a Windows version … Seems like I won't be using it much, coz my primary OS is Linux. (hey, please spend the last mile effort for the sake of Linux & MacOS early adopters…)