Sentience on Demand, as an Online Commodity

Charles Ostman

Senior Fellow - the Institute for Global Futures

tel 510 549 0129 fax 510 549 9689

email charles000@aol.com

This is an abbreviated, edited version of a paper / lecture, originally presented at the AAAI - Computational Anthropology 97 conference. The full version of this and related writings are being compiled for future publication.

*Synthetic Sentience, the New Commodity . . .

*Quality of Content, Indicies of Cultural Health

*Rentable Avatars and Disposable Personality Entities

*Sensory Seduction as an Engineered Process, the Invokation of Rapture

*Avatars as an Enhancement of Interactive Environments

*In the Beginning . . .

*Avatars with Component Personalities

*Self Evolving Personalities: Virtual Emotion Constructors

*Knowledge Engineering as an Industry, and the Spawning of the "Next Generation" of Intelligent Avatars, Artificial Organisms

*The Virtual Shopping Mall example . . .

*Virtual Amusement Parks

*But Is it Virtual, or is it Real?

Commodity assets have traditionally been specifically linked to the relative value status of so-called "tangible" items, such as agricultural products, bullion, manufactured goods, etc. Tangible asset commodities, though once seen as the defacto standard of measure representing the very essence of any modern market trading arena, are giving way to a new realm of "virtual asset" based commodities.

For the first time in modern history, a potentially irreversable fundamental socio-economic system shift is about to take place, where the majority share of "tradable" commodity "entities" are specifically based on virtual assets. Online, bi-directional interactive environments, real time electronic commerce systems solely dependant on digital cash transactions, and artificial intelligence based "synthetic organisms" accessable as nodal entities on the global internet system are only a few of the examples that comprise this newly emmerging arena of "virtual commodities".

A tangential byproduct of this realm of development, representing a substantial percentage of the total potential market share of this virtual commodity asset based economy will be in the arena of entertainment "product". Indeed, the very definition of entertainment, as a "process", will shift away from static, fixed, "units" of prepackaged, "sealed" production content, and move into the domain of interactive, dynamic, user definable "entertainment events".

Synthetic environments, populated by user definable avatar entities which are the emotionally sentient extensions of the human users, will at some point in the future, represent the mainstream of "entertainment" content. Entertainment, in this realm, may indeed become redefined as edutainment, in which multi user, mutli-nodal interactive environments become the situational equivilant of entertainment, educational, even "virtual workspace" environments.

We're poised on the edge of an era in which certain traditional events and places-cultural norms-are about to be augmented or supplanted by the virtualized versions of their physical counterparts. Entertainment locations ranging from shopping malls to movie theaters will have analogs in the digital environment. Online entertainment content need not be limited to merely observing or participating with a predetermined series of events. The entertainment event stream itself has the potential of becoming an ever-self-modifying, reconstructive environment which spawns its own contents according to the activities of the human participants. Furthermore, on the internet, the limitations of geographical and temporal constraint essentially disappear, which allows a consistently high refresh rate of these live contributions.

Contact with other entities in these virtual environments will be the interaction between avatar representations of the users. Virtual avatar entities are defined for this discussion as the self-modifying personality extensions of the humans they represent in virtual telepresence domains. This allows a potentially wide variety of experiences that human participants can enjoy, learn from, acquire, and build upon for future social and entertainment events. Aside from the exploration of virtual worlds, interaction with other avatars, participation in self-modifying, continuously evolving adventure environments, and other forms of entertainment, even something as mundane as visiting a shopping mall takes on an extraordinary new dynamic in the emerging virtual environments.

Quality of Content, Indicies of Cultural Health

Much speculation has occurred in recent times over the potential harm, and possible cultural degredation that may occur as the influence from violence and other negative content offered in current broadcast media affects the participatory audience. This is particularly noted when observing young people interacting with typical "shoot 'em up" style games, where predatory or destructive activity is the primary objective of "successful activity" within the environment. This, however, does not necessarily need to be the case. In fact, the exact opposite is likely to become the case, as the diversity of content is enabled by the enhanced sophistication of the synthetic enviroment development tools. In other words, "traditional" interactive computer games, which are for the most part variations of the same type of flight simulation or battlefield combat environments, are relatively "simple" enviroments to construct. In these realms, the "rules" of engagement, and the consequences of interactive event cue driven logic structures, are very orthagonal, and predictable.

By extreme contrast, synthetic organism constructs, with sentience-like properties and behavioral attributes which can be user definable, provide a vastly wider array of "social engagement" activities on the part of the user. It is at this point that the potential user base suddenly expands far beyond the narrow range of mostly adolescent male youths that have been tradtionally associated with "computer games". In this new arena of synthetic environment applications, the "game" becomes a virtual "theater" event. The potential audience appeal to a vastly wider demographic user base, with the market potential of online advertisng and marketing revenue directly coupled into these immersive "virtual theater" environments, is a concept very much appreciated by the developers currently involved with creating this type of content. It cannot be stressed enough, that even as this chapter goes into print, quite an extensive collection of serious, commercial efforts are already underway, developing synthetic environments, avatar based "character entities", object oriented media content, and other components relevant to this production process. Some examples will be cited throughout this text. Indeed, the very concept of thematical composition and script writing, as a process, is no longer described by a rigid, predetermined "story" structure. The challenge facing new content developers is not so much in the technology, which is already accelerating at a robust pace.

The real challenge is in the redefinition of the "story" itself. A writer is no longer faced with crafting a script, but rather in crafting the object oriented "personality organelles" of the characters within the approximated boundaries of a self evolving "story event". This is a concept that not all writers of the "near" future will be able to readily adapt to. Somewhat analogous to the transition from silent films to films with sound, a subset of the creative writing community will adapt robustly to this new paradigm, and perhaps many others may not. On the other hand, an entirely new artform of expressive content will emmerge from this realm.

Those who do find this new domain compelling, and begin to explore the almost infinite variety of potential "situational variables" that can be constructed into interactive, dynamic, synthetic environments, will find perhaps the greatest creative potential yet available to the modern "writer".

Rentable Avatars and Disposable Personality Entities

At some stage of development the concept of rentable organisms with root personality attributes becomes a marketable commodity.

A collection of ready-made graphical avatars and reusable personality elements could be available to the human user. Entire catalogs of personality features would become part of the "product line" of online synthetic environments. There will be a demand in this emerging world (both virtual and real) for libraries of characters with various forms of embedded functionality.

Deployment of "sentience on demand" will likely extend far beyond the perimeters of entertainment or recreation, and become a component of knowledge engineering and data mining. Much business today relies on the quality of the information base it can extract meaningful content from, as its primary weapon in maintaining a competitive edge. The ability to construct and deploy rentable research agents with quasi-sentient properties to "ply the waters" of the global information database will be viewed as a strategic resource, with an infrastructure evolving around this provider service. In its eventual implementation, an object-oriented personality constructor system, with user definable synthetic personalities in a dynamic, interactive immersive environment, will be common. These characters will possess the defacto equivalent of emotional sentience, and are the interactive audio/visual artifacts of self modifying code structures.

The artificial intelligence developments that much of the public has seen in the past isn't remotely comparable to the current genre of this domain. Recently, forays into this theoretical realm are being translated into functional systems (both hardware and software) that can provide a type of "synthetic sentience" as an avatar-based, virtual environment entertainment application.

Sensory Seduction as an Engineered Process, the Invokation of Rapture

Nature is often inherently seductive, and thereby implicitly interactive. The shape and color of a flower, the composition of a landscape, the ethereal kinetics of grass billowing in the wind all have alluring aesthetic appeal to humans. Form, shape, color, composition, as they occur in the natural world, are sometimes "engineered" to be as alluring as possible, to solicit the attention of the observer. In many cases, the purpose of the design is apparent, as a practical means to an end. The collective attractiveness of an entire natural environment, in fact, provides an avenue for its protection and preservation.

The seductive potential of an environment should not be underestimated. The degree to which rapture can be invoked is a measure of the effectiveness of the environment. As a means of sustaining interaction within the environment, even to the extent that leaving the environment represents a "sacrifice" on the part of the user, aesthetic features can have as much importance as the virtual organisms that may reside there. In a world where humans have the option to eschew the natural environment, there will be a need for ever more engaging synthetic environment interaction.

The paradigm of stimulation, pursuit of aesthetic content, and interaction with, will transition further away from the real world, toward the meterable commodities of the synthetic world. This is the threshold point of a synergistic co-dependence between the internet, with access to the synthetic world, and the human population drawn to it for interactive stimulation. This is a challenge, perhaps even a threat to today's cultural norms. But it's also a source of spectacular potential in aesthetic content development for virtual environments of the near future. In essence, the range of forms available in the real world, even in the most prolific rain forests, are a "fixed" value. There is a limit to the known range of available realms and their respective content here in the real world. That limitation disappears in the virtual world. In the future, the exploration of myriad aesthetically compelling synthetic environments will be a viable genre of entertainment.

Avatars as Enhancement of Interactive Environments

With the rapid emergence of multimedia as a means of providing access to products and services, the concept of "agents," which has been in existence for years, is evolving into "avatars" as a mechanism for enhancing the interactive experience of the user. Perhaps of greater importance to the creators of these media environments is the bi-directional data stream generated by an ever increasing and diverse user base.

This process extracts demographic data about the activities, reaction patterns, and behavioral attributes of users, in response to cues provided in the media environment. An important aspect of this process is to be able to "learn" the user's activity profile, as an indicator of their emotional response to various event cues embedded in the interactive experience sets inherent with the synthetic environment, without having the user actually being concious of this process as it occurs. Indeed, the more "hidden" this demographic data extraction process is, the more likely it is to generate a truly "accurate" interest and emotional response indicator of the user's participation in the environment.

This is where the "real" value is. In other words, the more robust the user driven "learning" process is, the more engaging of an environment can be dynamically constructed to create a uniquely compelling experience for the user. This dynamic process therefore has the potential of generating a demographic database that helps the designers of these environments create more compelling and seductive content. However, an even greater potential exists to use this data stream to influence the behavior and "evolution" of various synthetic organisms, and their synthetic personality attributes.

In the Beginning . . .

The concept of agents in general stems from the early days of attempting to create intelligent, automated mechanisms for contextually searching large-scale databases. The range of functionality, methods of deployment, and complexity of task handling vary widely. However, some common guidelines apply to most forms of agents. A "traditional" agent is the core of a search engine, that can search or query for a range of content of particular interest to the user.20 The more sophisticated the agent, the more approximate a definition of desired content can be. The ability to approximate values, as a test for recognizing a valid data set, is the beginning edge of intelligent agent features.

Some early versions of these agents employed versions of fuzzy logic, which is designed to respond to relative "nearness" of values, rather than to absolute values, as in more traditional logic structures. The most sophisticated search engines are those which can "learn" on their own from repeated input stimuli, and therefore develop an affinity factor for a particular range of a content type. Such intelligent agents often rely upon self modifying and self organizing code structures, which can in turn compose their own search criteria "on the fly" as a dynamic process.

These types of agent components tend to use neural net-like features, and often adopt behavioral characteristics that resemble the activities of an artificial organism, or "avatar." Avatars of today are not independent, self modifying entities with the virtual equivalent of sentience. That's not to say, however, that this is not possible. It's from the idea of "agent construction" as a process, and various related hybrids, that the precursors of synthetic personality entities become apparent. Currently, there are several projects in which synthetic personality engines and experiential environments are being explored. What's of interest is the diverse of range of strategies used to pursue this goal. The concepts offered here represent the implementation strategy I would use. This is by no means intended to be inclusive or detailed, but rather to provide a conceptual overview.

Avatars with Component Personalities

An important design feature in the development of an intelligent avatar is the ability to directly influence the continuous dynamic "reconstruction" of the avatar. This process heuristically renders a continuously evolving "virtual character" that can reconstruct itself to effectively respond to trauma, an opportunity for gain, or a change of conditions within an environment. In that sense, an avatar could be viewed as a near-realtime adaptor engine. In the design of such adaptor engines, the idea is to construct a personality with a core set of behavioral attributes, around which would be adaptor mechanisms.

Somewhat like the concentric layers of an onion, a collection of behavioral features could be attached and rearranged in layers of functionality to create the complete initial avatar. The layered subcomponents, or personality features, have both inter connectability and exchangeability within their class, and within their layer. This is the framework of a clustering model, as it's referred to in traditional neural net-like structures.

The personality subcomponents, clusters, and layers would be user-selected and user-organized, then gradually become dynamically self-organizing. With good software design, clusters of clusters of diverse (and dynamic) personality subcomponents would eventually begin to approach the functional complexity and specialization of bodily organs. Some developers currently even refer to such pieces of code as "organelles." In a personality constructor engine, the arrangement of the subcomponents, at the outermost layer, forms the essence of the personality. The number of layers, components, and the diversity of functionality required to render a truly heuristic model of a synthetic personality is a matter of debate. The object here is to merely outline the protocol representing this process.

Self Evolving Personalities: Virtual Emotion Constructors

The object-oriented personality components, with their characteristic behaviors, can be re-arranged and interconnected to form a personality matrix specific to the user. Personality component libraries would be available to a user who wishes to construct a "root personality" for their avatar, over which they would have initial total control. However, just as with evolutionary personality growth in humans, the emotional components of the avatar personality would tend to generate their own nuanced predilections over time.

This is where the boundary of emotional sentience begins to form. The behaviors of an artificial personality wouldn't be confined to those that were originally assigned to it. The human creator of the "virtual personality" would have a decreasing need to directly control the avatar's behavior. Conversely, the avatar develops increasing independence, perhaps ultimately even an awareness of "self" over a series of self-modifying personality iterations. Would the potential for rebellion exist here?

Perhaps this is the "fun threshold" of virtual environment adventure worlds. How much more stimulating a game, exploration environment, or interactive entertainment event would become if the personality attributes of the virtual characters, including your own, were independent, argumentative, and changed with experience.

What if sentience could be "hired" on demand, as an avatar presence? People today live vicariously through the lives of characters on their favorite TV programs. This process takes place in a real world where human interaction is still common and continuous. However, as bi-directional connectivity increases, and physical co-existence gives way to virtual co-existence, the market for enhanced, social virtual worlds will grow. Such worlds will become a key entertainment commodity of the cyber realm, one with complexity and richness of experience, both in the synthetic environments and in the emotional richness of the avatars that we use and encounter.

What would people pay for this-a way to experiment with or live vicariously through the experiences of their personal avatar? Consider a situation in which it can independently recognize and interact with other pseudo-organisms that it encounters in the environment you've chosen. The quality of such interactions would depend in part on the features assigned to your avatar. The more complex and adaptive, the more computing power is needed, and the higher the relative rental fee for access to these enhancem ents. Your avatar's behavioral reactions, and the patterns these reactions may be correlated to, would serve as a window into this synthetic environment.

An external observer of these patterns could be, in a sense, allowed the privilege of peering into this synthetic environment, to "see" what is about to happen, or at least have an enhanced predictive thread on which to base analysis. As an enabling technology for a new form of entertainment, as well as for market research, synthetic organisms comprise a very broad range of possible applications.

The Virtual Shopping Mall example . . .

Outlined here is a brief functional description of a user-specific avatar constructor engine. The application describes an interactive shopping experience enhancement in dynamic interactive virtual mall environments. In traditional retail outlets, the goal is to present to the visitor an aesthetic environment that displays the product in an inviting, stimulating and cost-effective manner. In a physical store, the environment is fairly static; as such, broad common denominators of decor and themeing need to be factored into the floor presentation of merchandise.

By contrast, in a dynamic virtual mall synthetic environment, the aesthetics of presentation and the attributes of the interactive experience are all dynamic variables that can be directly driven by the personality attributes of a shopper's intelligent avatar. As a representation of the shopper's personality traits, demographic features, and aesthetic and emotional preferences, the avatar's personality matrix dynamically shapes the appearance and affordances of the synthetic shopping environment. The end result is a dramatically variable, user-specific experience for each shopper, emanating from a common product database/artificial environment.

To evolve your avatar's personality, we need an initial stimulus event, from which the first layer of responses are extracted. It could be something as mundane as a questionnaire. However, in retail, the process could be made much more compelling, more like entertainment. If such entertainment events were embedded in the initial "visit" to the virtual catalog/mall front end, one could generate much more robust demographic data to "custom tailor" the virtual mall to the individual visitor.

As the user responds favorably to some synthetic mall stimuli, and less so to others, the end result is a structural template representing those features most likely to elicit a positive response on the next visit. This then becomes the feedback loop to generate a second layer of responses, which in turn can become the back propagation data stream to "guide" the user into interactively structuring their avatar. The more cycles of interaction with this avatar customization process, the more robust and user specific the avatar's characteristics become. They then become the template that shapes the hierarchy, presentation, and interactive features of the retail product presentations.

Virtual Amusement Parks

Virtual entertainment, designed to enhance and customize online shopping, is the logical precursor to establishing the components of complete amusement park environments in cyberspace. The shopping mall example cited above is merely a functional template for describing elements of what could constitute an almost infinite array of entertainment applications.

Synthetic, self-organizing, reconstructive environments are the obvious breeding grounds for immersive virtual theme parks. Furthermore, the range of aesthetic and interactive activities these environments can offer the vicarious visitor is extended by the features of other "intelligent avatar" visitors with their quasi-organic behavioral attributes.

Self evolving "virtual terraforms" are envisioned in this model as the landscapes, terrains, and structures that can be visited and interacted with. In my vision of this model, the virtual terraforms themselves are as dynamic, and potentially organic in functionality as the avatar entities that visit them. The synthetic environment itself can become an integral part the entertainment content. Even though at this time, there are severe limitations of processing resource, bandwidth, all are improving. Evidence of the rate of change, and what increments are at hand, are considered in this discussion.

In the beginning . . .

In the beginning, out of the primordial soup of creation, there was only a relatively small, eclectic collection of UNIXoids, confined to their room-sized CDC, Honeywell, and IBM mainframes. From this early beginning emerged the dawn of the Internet, a sparse network of thin tendrils of communication links spawning from a few score computer centers and research facilities. Establishing a connection to the net was time consuming, esoteric, expensive, and anything but convenient.

Roughly three decades later, many tens of millions of computer users worldwide, with more processing power on their desktop than most mainframe users of those earlier days could imagine, can plug into the "net" in minutes. Indeed, many of today's "newcomers" weren't even alive when the primordial soup of the global net was forming, and have no concept of what the early days of computing were like, let alone what the average UNIX user might need to know just to be fluent on the net.

* Knowledge Engineering as an Industry, and the Spawning of the "Next Generation" of Intelligent Avatars, Artificial Organisms

A trend is now emerging, where the sheer volume of information that could be perceived as being relevant, or even crucial, to the human participants seeking such information are beginning to rely upon automated, self modifying contextual "search engines." In other words, reliance on an artificial intelligence based, contextual search system is no longer an esoteric luxury. It has become a necessity. As explained earlier, developmental evolution of so-called gobots and knowbots was the "primordial beginning" of artificial intelligence driven search engines.

These were designed to be "trained" with a basic set of contextual parameters. Once "let loose" out into a network, they attempt to establish their own data streams and affinity factors to find whatever it happens to think might be interesting or relevant to its initial training " template." A training template is not a rigid parameter set, with an absolute match requirement to satisfy a logical "yes" value when an exact word, name, or other identifier tag is found, but rather an approximation value that represents a logical "similar" value.

The "relative similarity" value itself can be a dynamically modifiable feature set, so that a consecutively broader search context can be established, and redefined as the knowbot searches for, and finds, items to which it has established an affinity. Subsequent artificial intelligence search engines were designed with neural net like process elements. In search operations invoked with these newly developed types of systems, every instance of a successful approximation match strengthens, or magnifies the robustness to which another example of a similar stimulus will be responded to. In this way, such a process engine "learns" from its experiential exposure, and in the process builds up its own content addressable memory region. In other words, the longer such an "entity" is allowed to reside at a node and actively scan all data streams or log on to access nodes looking for something it "wants", the better it gets at the process.

Hence, the business of developing ever more sophisticated data searching, contextual linking, and "intelligent" processing has become an industry in its own right. In fact, for many businesses, the ability to have access to "knowledge engineering", as a resource, is as crucial to its operation in a competitive environment. It is the evolutionary development of knowledge engineering tools and techniques that is now spawning the next generation of tools, and resource base, with which to begin to create the "intelligent avatars", synthetic personalities and organisms, and the virtual environments they reside in.

Human behavior, and interests, are in fact being shaped by the structure of the internet. People are no longer limited by geographical constraints as the primary determinant feature in their activities or interests. The concept of residing in "cyberspace" as a daily activity, for leisure, social gatherings, political interest activities, education, etc., is considered "normal" everyday activities for a very rapidly increasing percentage of the human population. In a subtle, yet irreversible transition, the internet is becoming embedded in almost every aspect of human activity. As the embedding process penetrates ever further into the strata of daily human activity, the resulting symbiosis is enhanced to a status of perceived requirement by its users.

Restructuring the "virtual terraforms" of earth's inhabitants is being driven by the ever increasing growth rate of the global connectivity grid. It is the synergistic codependency between the global internet system, and the human population that symbiotically connects to it, that is at the very heart of this process. The concept of "entertainment", a fundamental requirement for the well being of any person attempting to function in the existence matrix of daily life, will likely change as radically as the technical and socio-economic components of the new age at hand.

It is in this arena that the internet plays an extraordinarily central role, will begin to take on truly organisitic behavioral attributes on a global scale, and drive the very essence of human behavior and motivations in the near future. Interaction with other individuals may, or may not, necessarily require "real" organic contact, but rather the enhanced, sensory illusion of an experiential event stream. It is in this cyber resource base that the components of an individual's existence matrix will reside, and to which the authority of access and control are dispensed.

Experiential technologies, as they become accessible via the global connectivity grid, are in essence the interface components for creating its organic, symbiotic characteristics. The enabling technology processes spawned by the implementation of the global internet system are rapidly converging into a new "existence matrix" which the human population will learn to adapt, and evolve into. At the heart of this realm is emerging an entirely different definition of what constitutes "entertainment", and by what mechanisms a sense of recreational and aesthetic fulfillment can be experienced.

The great dichotomy of this new realm is the series of tradeoffs, gains and challenges that this domain will present to its current and future occupants. The potential for a seemingly infinite variety of choices of entertainment content, both as a singular experience for the individual, and a collective experience for "virtual communities", can seem to be a tantalizing allurement to some. Perhaps to many, however, this experience may also have the potential of being as daunting as a mirage.

But Is it Virtual, or is it Real?

As of this writing, various "working models" of the realm described above are beginning to emmerge into the "real" world. This is, however, a very rapidly "moving target", somewhat like attempting to hold water in a fishnet. At this particular moment in time, there are still a variety of "practical" limitations, some of which are strictly technical, and others which have more to do with the "inner machinations" of key players in the industry, which will set the pace of ubiquitous immersion into a truly seamless, user friendly internet connectivity grid populated with mhyraid avatar entities and synthetic environments.

The major components of establishing this "new realm" are increased throughput data rates for internet connection, most likely to be accomidated in the near future by so-called "cable modems". In this modality, cable TV providers will be providing internet access via the much higher bandwidth connection available through their coax cable system than through standard phone system wiring. This coincides with the rapid emmergance of the various digital audio and video compression/decompression codec chip sets, and also the GLINT (graphic library integer math) chip sets which allow for near realtime 3D rendering on the local users' computer or "settop" box.

But these are only minor, momentary access enabling "cogs" in the wheel of the much larger scenario. The real issue is the rapid deployment of agent and avatar based "entities" and environments on the internet, the development tools available to the general engineering and asthetic content development community, and the drastic improvement in cost/performance characteristics of computing platforms required to support the environments and their "contents".

Computing capacity, as a general resource, will shift away from user dependant harware and software, and much more towards server dependancy. In other words, the "power" is at the server node, where supercomputing capacity capable of supporting the extremely complex computational task handling associated with "synthetic sentience" rendering, dynamic environment deployment, and related processes are transparent to the individual user. A particularly interesting example of a "low cost", stackable, plug and play supercomputing "module" that could find its way into the online processing node application realm is the ICE system, originally developed at MIT, and now available as a commercial product.

This is a 64 SHARC CPU chip board capable of spewing out a continuous 7 Gflop computational performance capacity, and is specifically designed to be accessable as a node in a multi-nodal, parallel processing environment. But, discussing "hardware engines" at this point is abritrary, at best, since there are multiple other developments and platforms poised to enter the "market" as this chapter goes into print. But, are there agent and avatar based applications and environments already "flourishing" on the internet? The answer is, it would be a far too numerous, if not rapidly changing list to simply recite here, but there are some particularly compelling examples which serve as reasonably relevant "marker points" in the current application development realm.

In terms of a practical, revenue stream bearing "agent enviroment" application currently online is the "FireFly" site, created by Agents Inc. Here, the initial entry of the user through the loggon process outputs a demographic "data template" back into the agent engine, which in turn utilizes this data stream to send various media titles and "samples" back to the user. The user base gets to have access to a variety of "fun" activities, including live "web chat" sessions, and the paying advertiser clientelle gains access to a very selectively targeted, potential "customer base" for their music and other media content product line.

As a business model, this is fairly representative of the type of interactive, immersive, realtime demographic "targeting" modalities which will undoubtably become the new narketing platform of the virtual environments of the near future. Already, there are a number of examples of "agent environment" domains, in which intelligent agents actually interact with other agents independant of the human "monitoring" of their activities. Although a number of academic development sites have pioneered this realm since the late '80's, commercial vendors, such as Gereral Magic, are now offering devlopment tools and applications to specifically support this type of agent interactivity "knowledge" processing. Another such example which bears scrutiny is the MCC InfoSleuth project, which is perhaps one of the most aggressive attempts yet to create "spawnable" agent interactivity as a process. Indeed, this emmerging realm of collaborative "information discovery", pattern set recognition, and so on, is almost as complex of a topic as the realm of truly "organic" process engine systems utilizing neural net and genetic algorithm structures.

It is only the purpose of this writing to offer an overview of the general concepts, and application examples, of intelligent agents, avatars, and "synthetic environments". In terms of actual dynamic avatar populated 3D environments on the internet, as in "VRML worlds", the total number of actual VRML content compatable sites is relatively low, but this is a very temporary artifact of the current technology commercially deployed. Of course, there is the Worlds Inc. site which is by far the most well known of these examples, but others are rapidly emmerging.

A particularly compelling example is the work being generated at the Construct site, through which I recently "visited" a live, online, avatar populated "trade show" event taking place in Tokyo, Japan. For anyone with web access, however, even a brief search of the Lycos or Yahoo search engines yields up literally hundreds of agent and avatar technology development sites, various vendors supplying VRML browsers, content development tools, and so on. The "real meat" of this domain is in the development of neural net and other similar "organic" process engines capable of rendering self modifying, self adapting "behavioral entities" which would be at the heart of these synthetic organisms and their dynamic environment component systems.

As of this writing, there are at least four specific examples I am currently aware of in some form of development, and a number of others are "rumored" to exist, which are poised to enter into commercial applications development. One example of an avatar populated "behavior engine", which offers a glimpse into this realm, is the Improv system, first demonstrated at SigGraph in 1994 (Orlando, Florida). Unlike other synthetic environment domains populated with self evolving organism components, such as the Avolve "interactive plant growing" system, Improv actually demonstrates behavioral characteristics assigned to "humanoid" characters which can be user defined as an interactive "theater" environment.

copyright © 1997, 1998Charles Ostman.