Muq Plans

This short chapter is included primarily for legal reasons.

Ideas I have considered for inclusion in future releases of Muq prior to March 1, 1997 include:

• Improvement of the current in-db transparent world networking support, possibly to the point of full network transparency at the application user level: Users seeing a world seamlessly distributed across multiple servers at the LAN and/or WAN level.
• Radical improvement of existing semitransparent inserver networking possibly to the point of full network transparency at the application programmer level.
• Implementation of three-dimensional graphics worlds of the VRML or OpenGL sort, based on added inserver graphics support and/or a separate graphics coprocess.
• Implementation of such worlds based on heavily procedural world definitions, with actual geometry, textures and/or other graphics elements generated on the fly rather than stored statically.
• Use of a single underlying world definition, static or procedural, to drive both VRML/OpenGL style interactive graphics, and also photorealistic rendering, most likely at noninteractive rates.
• Use of cone-casting or the like (as opposed to classical raytracing) to allow relatively efficient rendering of scenes in large and/or complex virtual worlds/universes.
• Use of "textures" or similar facilities as caches to re-use parts of the scene rendering, perhaps those changing most slowly or most distant: One might render distant parts of the scene into texture buffers and render those parts quite efficiently using many fewer polygons than would otherwise be needed, recomputing the cache textures and polygons periodically as required by changes in the viewpoint and/or the background itself.
• Extension of such partly or wholly procedural world definitions to describe very large universes, perhaps having billions or trillions or more of galaxies, each perhaps containing similar numbers of worlds, each perhaps containing similar numbers of nonplayer characters or other inhabitants, and/or extension of world detail to very small scales, potentially atomic or smaller.
• Similar construction of universes of other geometry but of similar scale.
• Extension partly or wholly procedural world definitions to time, such that one can potentially travel freely through both time and space, and find substantially the same scene at a given spacetime coordinate on each visit.
• Extension of such procedural world definitions to include such nongeometric world aspects as language, culture and literature.
• Use of permutation functions, spline functions, cyclic functions or the like to allow motion of large numbers of objects in large and/or complex procedurally defined universes while retaining the ability to render local scenes relatively efficiently, in particular by locating the set of objects within a restricted volume in time more like O(1) or O(logN) than O(N).
• Construction of virtual worlds mirroring actual pre/history and/or geography.
• Political structuring of large virtual online communities as heterogeneous trees, lattices or graphs. The B-tree might be particularly useful. "CommuniTree" might be recycled as a name for such a structure.
• Partial or wholly automating some administrative, legislative, judicial and/or police functions (or the like) in online virtual communities, such as by automating some portion of something like Roberts Rules of Order, or providing copbots or such.
• Extending the mud concept to a heavily or wholly HTML (or similar rich text format, or perhaps VRML or such) substrates (as opposed to conventional plain text), perhaps via MIME or the like.
• Using the web as an interface to the mud, possibly via Java-based mudclients: mudserver as webserver.
• Similarly, using web facilities to store/generate some or all of a world definition (or avatars or such): mudserver as webclient. E.g., virtual libraries could be based on online resources such as the Gutenberg Project, either delivered straight or perhaps processed so as to give (say) the impression of a large library in an alien language.
• Introduction of neural-net style computations, perhaps to provide better nonplayer characters or 'bots; More generally, application of AI concepts and tecniques virtual worlds and their contents.
• Use of hash functions as implicit representation of neural net interconnects, perhaps to economize on time or space.
• Use of long scalar vectors to represent knowledge more robustly than do classical boolean AI approaches, the scalars in question being anything from bits through bytes and integers and floats to complex or quaternion values, vector length optionally scaled to unity or a similar standard values.
• Similar use of extremely long sparse vectors.
• Use of dotproduct (or similar, in particular linear) distance metric on such or similar vector spaces, rather than the classical Euclidean distance metric, perhaps to emphasize commonalities over differences in the comnparison.
• Semi/balanced tree storage of such vectors, allowing storage and/or retrieval of vectors in roughly log(N) time, and/or retrieval of close match(es) to a query vector.
• Use of principal component analysis or the like to improve such semi/balanced tree storage.
• Representation of the short-term memory of an automaton as such vectors, optionally stored in such trees.
• Regular storage of such short-term states, perhaps in such a tree, perhaps to provide automatic short-circuit of infinite loops via recognition of recently visited states, or to otherwise guide computation in the automaton.
• Understanding of relationship between data compression and pattern recognition, in particular informed by information theory.
• Understanding of pattern recognition as procedural data compression, and in particular the quantitative application of information-theoretic concepts and analysis.
• Application of such procedural data compression / pattern recognition to knowledge bases and/or data bases, and in particular to such vector databases -- and understanding of storing experience to memory as an active computational task involving search and/or optimization, rather than a passive recording task.
• Construction of an automaton using above-sketched sort of compressed and/or processed vector db of past states under dotproduct metric or the like to partly or wholly guide state transitions, perhaps to achieve flexible and/or robust computation based on past experience.
• Induction of "midpoint" subgoal production rules in such a context, to allow long-range plan construction in logarithmic rather than linear time.
• Introduction of hardwired goal/happiness functions to supply motivation to such or similar automata, perhaps as distinguished input values in the input vector.
• Encoding of "motor" (effector, in the broad sense) outputs as distinguished elements of the state vector, fitting output naturally into the central state-transition mechanism of the automaton.
• Application of such automata to virtual worlds and/or communities, possibly as nonplayer characters, impartial authorities, help facilities or the like.
• Aggressive expansion of the existing exception-handling system into a facility for routinely inspecting and guiding large and/or long-lived computations, even in the absence of anything which would normally be considered an error or exception, perhaps in the context of adaptation of Muq to be a high-level framework for organizing such computations.
• Introduction of constructs midway between conventional avatars and nonplayer characters: A user/visitor can "occupy" them and see the world through their eyes, but has sharply limited control over their actions. E.g., the construct might be driven by nondeterministic code defining what its allowable ("in-character") behavior is, with the user allowed to control/influence the nondeterministic choices made.
• Understanding that any and all free parameters in virtual world construction may be used symbolically as well as literally. For example, any visible property whatever may be used to distinguish a given object from its surroundings, drawing the user's attention to it if the world designer so desires: Hue, saturation, intensity, level of detail, motion, specularity, opacity -- anything.
• Understanding that apparent information flow is in general minimum after initial noise and redundancy are removed from raw user input and rise steadily thereafter, and hence that to maximize realtime multiuser interaction rates in the face of communication bandwidth bottlenecks, that the local world context needs to be duplicated on each user-local machine, and communication largely limited to propagation of lightly processed user input from other users in the vicinity.
• Understanding that in general the Net is effectively special relativistic: "Simultaneous" doesn't exist except for events physically close to each other, and software designs must recognize this.
• Defining user identity directly in terms of a public key: That user -is- everyone who can sign messages relative to that key.

Go to the first, previous, next, last section, table of contents.