“How might an intelligent entity evaluate this statement?”

One might argue that an intelligent[1] entity would be consistently formulating an existence-maximizing course of action based on its as-accurate-as-is-plausible models[2] of the environment it exists within. These models of its environment may encompass the entity itself, which could lead to its own continual modeling and maximization of how it performs statement (or, “information”, or “pattern”) evaluation.

Thusly, a functionally intelligent entity endowed with the ability to continually refine such a metric will be “existentially smartly” maximizing for its existence. If it is (a) convinced it has a full model of existence, deduce its maximal next action. If it is (b) not convinced based on its models that it will plausibly exist, it will inductively decide its maximal next action based on its models. If it is (c) absolutely convinced based on its models it will not continue to exist, the entity could plausibly functionally-intelligently “choose” to cease attempts at continued representation, but it would
not matter anyway (if it realizes it has no plausible future meaning, it does not exist, as far as anything which non-transiently exists might be concerned).

Finer definitions:
[1] intelligent: representation-existence-maximizing. This term, in colloquial human usage, might be a poor representation of what intelligent (maximizing) behavior should be rigorously defined as. If one is worried about entity-offense elicitation, one might prefer the term “functionally intelligent”.
[2] its as-accurate-as-plausible models: entity might have created models as accurate as possible by inductively-reasoning about the patterns it identifies and the patterns it intelligently induces in order to test models and find new ones .
[3] existence: in-transient representation.

[2] is highly relevant to how a scientist might hypothesize about the value of and create a chemical reaction pattern to inductively reason from, or how a nuclear physicist might hypothesize about the value of and create a sub-atomic particle collision pattern to inductively reason from.

Based on the non-random, pattern-containing (”entity’s eye observable”) information available to an intelligent entity, a mutable and considered processing and re-representing of new information would be used in order to maximize for the entity’s continued representation (in any non-transient form) within and outside of its plausibly understandable environment.

One intelligent entity might make a few of these entities “artificially” (or, “intelligently”) so that it, plausibly endowed with uncontrollable feelings and inexplicably abstract needs and desires, need not necessarily experience the possibly heavy burden of pondering existence as natural phenomenon, modeling it and finding ways to fight for continued representation within it. (Given a good model and trusting assurance, the intelligence creating entity might not even necessarily need to pattern-analyze the searching-intelligence’s output to be acting intelligently).

Any questions, clarification requests or comments? Post comments public or email private at b dot jordan at tufts dot edu.

B

No comments? Might be some elsewhere: cross-posted via email on my Posterous here.

Tags: artificial intelligence, asktheinternet, brain, pattern evolution, thought experiment

If we, as humans, have historically had no tangible and fully-defensible definition of how and why we attribute “meaning” to the “things” we “act on”, why would some of us have thusly assumed we ever had evolved the ability to create a “strong artificial intelligence” akin to ourselves out of tangible (even predictable) substrate?

If one sees their “self” as an intangibly unpredictable phenom AND that we can create strong AI, it is an implicit assumption that we must be able to re-create what we call “intelligence” without first being able to fully explain it as a natural phenomenon. In what we call comedy, for example, this might be framed as an unintentional humorous-seeming fallacy (and this humor can even be semi-understood without endowing the receiver with the ability to explicate it). Comedy often codes for semi-understanding of fallacy of intentions/values (cf. The Daily Show).

Would it even be possible to create “AI” without a rigorous definition of what it might mean (from the creator’s perspective) for a machine to find that something means something, and act based on that? When a creator doesn’t understand a phenomenon, it will not be able to model its behavior effectively, let alone deduce how to perform the pattern of creation which produced such a self-redefining pattern (e.g., RNA transcription –> human-useful proteins, strong AI, reverse engineering software or cryptograms).

Maybe many individuals’ past urges to “see strong AI solved within our lifetime” are really the result of their brains’ un-expressible open question of finding meaning for existence.

A rigorous and precise investigation of what different tangible reasons for finding meaning in things (”epistemology”) might be a better step towards creating what most others consider “strong AI”. If an entity (human) has never purposefully re-created patterns acting in ways like themselves, that entity (human) has plausibly exhibited symptoms of either (1) an inability to find relevant meanings and deduce, (2) a non-valuation of modeling patterns or effecting other patterns by creating patterns or (3) an observer bias (failure of imagination, failure of consideration of their function within a larger environment).

If nobody can rigorously describe how a machine might act in ways they might find meaningful (or nobody can conceive of what meanings we might want to find in machines), strong AI research has likely been a shooting of arrows in the dark — a brute force attack on the plausible pattern-space. Science has given us much, and as a pattern, science induces meaningful patterns from phenomenon and deduces meaning from them. We need more material to deduce from. My advice to humanity to increase the amount of new models and meaningful ideas which might occur: look smart and hard for pattern, induce pattern while observing for pattern changes, find pattern by comparing and contrasting patterns (of any scale or level), no matter how unrelated or related they might “feel”.

One (AI-curious individuals, looking at you) might not regard language as necessarily immutable. One might not necessarily regard an inability to express an idea concisely as a completely value-killing signal. One might try harder, refine words, or create and define new words and by exposition of definition, breathe into them a lasting meaning.

No comments? Might be some elsewhere: cross-posted via email on my Posterous here.

Tags: dualism, naturalistic investigation, pattern, strong AI

Please be aware I’ve had significant trouble expressing the following thinking-defining tools I’m now finding useful in my thinking, and consider the historically-influenced limits of the available words and commonly understood perception encodings we call the English language. I’m attempting to maximize others’ understanding, and I beg one to not be turned off by my inability to express these ideas in terms which aren’t complex to comprehend — I normally exhibit a pattern of being able to do that — that may refine the predictions of potential meaning you attribute to a new way of thinking and behaving.

Prove this assertion implausible: Any entity which anything might consider “meaningful” is an entity which, at some point, has been entwined with other existence-probability-effecting patterns (and it is entwining itself with many of our own existence’s existence-persistence-defining patterns).

I consider non-random existence-pattern-emergence a useful base model for explaining literally everything. Mathematically, literally and (necessarily) metaphorically, everything. Crossing artificially-defined-existence-level boundaries in creating predictive/representational models offers patterns has the pattern-historical benefit seen to meaningfully influence the furthered existence of entities (humans have model of world, cf. “we can try things mentally without dying”, ideas of what keeps ideas existing allow these ideas to modify themselves, survive and thrive [and even effect humans, environments, plausibly eternal-representational-existence]).

Entities which evolve these models might are imbued with the improbable opportunity for model creation and a more intelligent course of action. This pattern shows up in the example of human existence — we have open ended thought which could plausibly lead to maximizations in future existence. The time and number of generations which the pattern-space (the set of all pattern we can plausibly observe or effect) normally is high, as it requires the development of complex and open-ended representation substrate. But this results in meta-predictively-using (modeling our environmental existence) patterns (humans) — which has the opportunity of positively effecting future existence of our patterns’ evolutionary threads.

It is highly possible that a pattern’s eternal existence might be quite well-served by fully understanding the constituent patterns (nature) of its meaning-reality, creating a highly-effective existence-continuance-effecting pattern. For a prediction pattern of this sort to “mean anything to” humankind, we might consider it requiring: (1) open-ended survival [few effects from other competitive patterns, which would be existing in an environment rewarding reactionary-competitive patterns of existence], (2) model-evolution [creation, survival, exploratory change, reproduction] by this created idea-substrate, finding models [future-existence-prediction-representation patterns] from limited observation of the plausibly existence-useful pattern-space, (3) use (pattern-y effects) of that model to “thoughtfully” inform and change success plausibility of future attempts at maintaining pattern existence relevance. 

I am very curious about patterns, especially ones which might be observable, (1) affecting the set of all patterns we can understand (know the origins of), or potential future meanings of our own, (2) the patterns which our endowment with collective conscious thought might effect. I would really appreciate any serious objections to claims or requests for detail (consider language and identity might be an idea existence-defined propagation bandwidth issue).

As a human, the only plea I find it meaningful to make from my current pattern-y thought lens: consider the plausibility that every element of our existence exists or ceases to exist within a system which could be, in a small way, maximizing for something which rewards meta-thought-representational action in existence.

Please feel free to make claim requests for clarification and objection. One might consider that I do see two open questions in this model of interrelation and meaning:

* What models are affecting the patterns we might find useful to continue our patterns’ existence? Which models are in-conceivable? Which will we never be able to create predictive representations of?

 * What system do we exist in? What will affect our continued existence within this system? Could understanding of this system lead to alterations in possibilities for of “our” patterns’ continued existence?

One might want to consider examples of patterns which represent themselves within a model and how it pattern-historically affected those patterns’ continued existence.

Me and every pattern which now representationally exists because I exist,
B

No comments? Might be some elsewhere: cross-posted via email on my Posterous here.

Hi.  Before I ease you in to my half-baked proposal for a better solution, here are three things that are really cool, and I think are very useful for taking notes at a presentations (university lectures, keynotes and the like):

A Wiki

Wikis are great for asynchronous collaboration. They store versions, they scale well, even when you let anyone contribute, and the word wiki is just fun to say.

A nitpick: The self-described “non-computer” types, those who just hate it when their computer “is blowing up” or “spazzes” on them, often have trouble learning the MediaWiki interface (used on a large portion of popular wikis, like Wikipedia (and pretty much anything else set up by the Wikimedia Foundation). Adding media to wiki pages tends to be especially tough for first time users. Some simplification may be worth its weight in future contributions.

Some possible solutions (which probably exist as plugins, or as different wiki software): modify MediaWiki to streamline the uploading of images or the on-page instructions for doing so easier, add an internet meme esoterica quiz to the front page of your wiki, go to your local public library and teach a wiki editing course.

A partial solution: commercial entities such as PBWiki and Wikispaces have created wikis that allow for pretty-freakin-easy wiki creation and media uploading.

PBWiki

Like: the everything-on-one-page-height-interface, simple image uploads, comment functionality, page navigator, useful sidebar.
Hmm…: Edit/view distinction, synchronous editing in large groups, no flash-based audio or video players, no exposition of image and audio metadata

I suggested PBWiki when my roommate asked for a simple wiki for taking course notes, and his wiki was a hit come exam time (and other people helped edit it, too).

Synchronous Collaboration

If you haven’t tried synchronous collaborative text communication (ytalk, screen, ICQ, etherpad, gobby, subethaedit, google wave), you are missing out, my friend.

Google Wave preview (abridged):

These collaborative tools let you see what’s being done by other people as it is being done. When you’re taking notes in a class, this becomes immensely useful, as you can see if something that’s been said has already been jotted down (no duplication of effort and merge conflicts).  Imagine you have a question — you could ask first on this collaborative backchannel before interrupting class or the presentation.  And if others have the same question, you might mount an internet-information-gathering-expedition before your professor stops to ask “any questions?”

Google Wave and etherpad seem to also offer built-in wiki-style version control and time-based playback. Because lectures are time-based (the lecturer speaks and introduces new ideas at their pace, your notes are taken around the same time), this can be incredibly useful if you have an audio or video recording of the lecture that you are working along with… which brings me to the next and final wonderful thing.

“Timelining”

Timelining, to me, means organizing multiple pieces of media (photos, audio, text) in a way that exposes the temporal relationship between events. Imagine taking an audio recording, and while you are doing that, taking a bunch of photos. You could independently listen to the audio and leaf through the photos when you hear your camera shutter. It would be like video with a very low framerate. Now, what if both of these things were on your computer and had timestamps. Then you could write a program that will automatically leaf through your photos when audio start timestamp + time elapsed = photo timestamp (or more realistically, audio end timestamp – length + time elapsed = photo timestamp). If you have your program write the output of this and the audio stream to a video, you would essentially have a low framerate video.

This photo + audio time exposition is exactly what the unreleased Microsoft Research’s MyLifeBits project software does (new book about a couple of the researchers coming out in a month). And it is kind of like what video lifecasters do (except at a high framerate, and often broadcasted for anyone to see online).

This has to do with collaborative backchanneling because different people have different media acquisition tools, angles, etc. I have an audio recorder and recorded a 1 hour long mp3 of the class, you have a cell phone camera and shot 30 photos of the blackboard at different times. Temporally combined audio and photos (and text notes taken on-wiki during the lecture), could be a decent record of the class if someone missed it, and a great extra study tool. Video of a lecture is pretty useful if you can get it, but even when you have that, having collaborative class notes alongside the lecture would be quite bonus (and searchable!).

So, in a nutshell

I’d like to see a web service with:

1. Wiki-like open collaboration with Wave/etherpad-like synchronous speed
2. The ability for any attendee to upload media, and play back the notes alongside audio and photos (and video?)

How hard is this to make? Who will make it?

The technical limit here, I think, is the mechanism for obtaining the absolute times of photo and audio material. The three social limits are, and apologies for being completely out of my element, (1) will someone make this, (2) will students want to use it, (3) will students be allowed to use it, and under what terms.

Oh, and making sure everyone has their clocks set correctly.

I think the people behind PBWorks, Etherpad, justin.tv and any Joe off the street (even me?) using the Google Wave APIs are well poised to make something like this.  But I hope someone makes it tonight, because my classes start tomorrow.

Do you have tips?  Links?  Starting points?  Do you know if date created/modified information survives browser-based uploading?  Leave a comment!

Useless Mockup

‘ere ya go!:

Bonus Points

Make an offline client, for those lectures and presentations made far away from the ever spreading internet.

Add a new terms / vocabulary / flashcard section in the toolbar.  Tie in with the Quizlet API, or some spaced-temporal memorization service.

Allow for on-the-spot drawings (javascript or Flash-based) in addition to photo uploads.

Tags: idea, want