[agi] Games and Funding

[A M]

--

Hello everyone,

I have been busy the past few weeks hence the lack of input. But I will chime in on the funding front. Firstly, I should introduce myself to those who don't know me (most of you). My name is Abdul Malik and I work for 1000 Planets Inc. ( www.1000planets.com ) a space development company, we are tyring to develop the necessary infrastructure to enable development of space i.e. launchers etc. One area of great benefit and importance (one of NASA's great hopes for cheaper, faster, better space missions) is A.I. and for us AGI. Now onto the body of the e-mail.

It is fairly obvious that conventional funding for AGI or AGI-related R & D is virtually non-existent. Most funds available for A.I. research hail from segments of the government i.e. the NSF, military etc. Moreover, such groups (with the exception of the military, which leads to ethical and moral concerns) are more sympathetic to narrow-based A.I. research. Narrow-based A.I. is not as grand as broad-based A.I. but it is also cheaper, relatively less challenging, more predictable, more definable etc. than broad-based A.I. Incremental stable and specific development is considered to be highly advantageous for all would-be sponsors of A.I.

Why invest millions in A.I. development when no one knows what "true"/ "real" intelligence let alone what architecture/working definition/attempt will lead to it. Hence the proliferation of narrow-A.I. in academia (where the funding authorities are ultimate masters) , business (for obvious monetary reasons) and government (better uses for tax dollars).

Now some have suggested that for certain applications broad-based A.I. will eventually become necessary and the A.I. community will be free to pursue the old dream with more resources, more will and more hope.

It’s a nice dream but it will likely remain a dream. Basically put, narrow-A.I. meets and will continue to meet the needs of most applications. Applications that require more AGI-type capabilities are few and far between and can be potentially be solved jointly by conventional information systems and narrow-A.I. and another intelligence : humans. After all, do you really require a voice-processing system to have creativity?

It has also been suggested that AGI-level A.I. components can "compete" with narrow-A.I. The underlying assumptions with this approach are that AGI-level A.I. can be reduced to some definable components and such components can be utilized in a predictable, definable manner. Pretty tenuous.

All of this assumes of course that AGIs can be "developed competitively" with narrow-A.I. That it is to say : AGIs can be developed within identical budgetary and resource constraints. Also highly tenuous.

I believe the realistic option, barring visionary, deep-pocketed investors/philanthropists, is to pursue the self-funding option. Moreover, I believe that the most efficient self-funding path is through employment of the general population as funders.

Industry is more supportive of conservative, immediate-payoff, predictable A.I. efforts that can only be, currently, met by narrow-A.I. i.e. the ever-present expert systems. Moreover, any attempt to create AGI-level A.I. would naturally result in very restrictive arrangements with the developers : "real" / "true" artificial intelligence would be considered a competitive advantage to be closely guarded.

A relevant example : 1000 Planets Inc. approached a number of insurance providers for the satellite industry approximately one year ago in efforts to recreate potential investors for an Orbital Maneuvering Vehicle (a "space tug boat" with spacecraft servicing capability). Such a spacecraft would reduce satellite transportation costs in earth-orbit while reducing risk and enable spacecraft repair and upgrading. However, following the interest we were bluntly told by one insurer that they would not provide funding as they felt that they would be helping their competitors. In a "cut-throat" industry where a single satellite loss can cost millions the insurers were concerned that by helping to reduce collective risks they were potentially helping their competitors. The other possible way we could have secured their funding would have revolved around us becoming a strategic partner/an "in-house" project/a subsidiary. Inevitably such a course would fail due to perceived costs i.e. technological risk, money etc.

The general populous is more supportive of A.I. development efforts - including AGI - than both industry and government. Some apprehensions do exist in regards to AGI however, they will eventually need to be addressed regardless of the funding avenues utilized. In addition, the general populous collectively surpasses traditional funders of A.I. in sheer funds.

An example : the computer/console entertainment industry (estimated at over 20 billion USD) exceeds the movie theatre industry and will soon eclipse the movie rental industry as well. The impressive growth rate (est. ~ 24 % annually) is predicted to recede but remain relatively high. Moreover, the entertainment industry collectively and computer/console entertainment industry in particular is relatively recession resistant (not proof but resistant). Most importantly, A.I. development is considered to be the main limiting factor as graphics continue to rapidly forge ahead.

The pros and cons of the computer/console entertainment industry for A.I. R & D have been discussed at length within the A.I. community.

Unfortunately, for AGI development there are a number of present barriers : current personal computing resources are insufficient ; ethical and moral questions exist (i.e. first-person-shooters) ; funding remains difficult to obtain ; deadlines and budgetary pressures ; immediate expectations etc.. However, I believe there exists an alternative.

Instead of attempting to develop AGI within the framework of the computer/console entertainment industry I submit that AGI development groups create products and services that generate revenue that may then be used for AGI development.

One idea that 1kp (our company nickname) is pursuing is a interactive, "game show" computer game (GSCG) utilizing synthetic/virtual actors as game show host and in a latter edition as co-contestants (the original contestant is the user). Dynamic, behavior-based animation with high-fidelity 3D graphics would create an engaging experience similar to "Jeopardy" / "Who Wants to be a Millionaire" but with eccentric characters i.e. a "Star Trek" "Q"-type character as host ?). Multi-player would allow for multiple contestants and the possibility of league tournament play. Questions would be based upon actual knowledge and not on mindless TV trivia.

Such a game could make due with a basic A.I. and could be developed within months. However, the revenue generated from such a game, we believe, would be quite high. Sufficient not only to create new titles or versions but also to fund AGI development and basic space development efforts i.e. spacecraft design.

An illustration : Id's John Carmack has developed and marketed graphically innovative games, creating several franchises in the process. Today he also has his own aerospace company (Armadillo Aerospace) and has entered the X-Prize.

We intend to create successive versions of the GSCG and other computer games/simulations each with better graphics, animation and A.I. until highly advanced levels are reached i.e. real-time movement, natural language understanding etc. Each version will lead onto the next and also fund a separate development effort : towards AGI.

Ben, we would like to collaborate with the RAI on this. The first version will best be served by a basic A.I. successive efforts would require more robust A.I.

Although we are in collaboration with the RAI on AGI (a collaboration that has previously been more in word than in action, we hope to change that) we are open to collaboration with all AGI groups in self-funding. We believe that multiple approaches to AGI is a good thing (Pei's mountain path analogy would fit well here) but only if they are properly attempted, i.e. enough funding for enough thought, and actually completed. Proper failure can lead to as many insights as proper success, in the end everyone benefits.

Look forward to your response.

PS : I just saw the number of e-mails about game A.I. I apologize if I have been redundant. I just got acess back to my AGI e-mail address. A lot of reading to do :)

Also my primary e-mail address is [EMAIL PROTECTED]

_________________________________________________________________
MSN 8 helps eliminate e-mail viruses. Get 2 months FREE*. http://join.msn.com/?page=features/virus

-------
To unsubscribe, change your address, or temporarily deactivate your subscription, please go to http://v2.listbox.com/member/?[EMAIL PROTECTED]