TELOS, A Language for Building Well-Structured AI Models
Proceedings of the Fifth International Joint Conference on Artificial Intelligence. Languages and Systems. Pages 917--1.
© 1977 International Joint Conferences on Artificial Intelligence
Abstract
- Travis, L., M. Honda, R. LeBlanc and S. Zeigler. 1977. Design rationale for TELOS, a PASCAL-based AI language. Proceed i ngs of SIGPLAN-SIGART symposium on AI and programming languages.
References were extracted automatically from the PDF and may contain errors.
Full text extracted automatically from the PDF; figures, tables and equations may differ. The PDF is the version of record.
TELOS, A LANGUAGE FOR BUILDING WELL-STRUCTURED AI MODELS
L. Travis, M. Honda, R. LeBlanc and S. Zeigler Computer Sciences Department University of Wisconsin - Madison Madison, Wisconsin 53706
TELOS is an attempt to provide powerful ab straction mechanisms and other structuring facili ties within a language that provides the special capabilities needed for AI research. TELOS in cludes PASCAL as a subset and also is implemented in PASCAL. A full description is available in [1] and [ 2].
Like most other AI languages, TELOS includes facilities needed for experimentation with large stores of general knowledge, tentatively modifi able and associatively referenceable, and with various planning and reasoning strategies. Howev er, in contrast to other AI languages whose design has focused on building in certain powerful high level constructs, the design of TELOS has focused on building in powerful abstraction mechanisms with which these particular high-level constructs, as well as numerous others, can be defined and im plemented with reasonable ease.
The usefulness of abstraction 1n all kinds of human intellectual act1v1ty has long been ob served, and programming is certainly no exception. Only recently, however, have programming languages begun to appear with features specifically design ed to facilitate abstraction of the several dif ferent kinds needed in the programming process, in particular, data abstraction and control abstrac tion as well as procedural abstraction. TELOS im plements a set of data , control, and procedural abstraction mechanisms specifically tailored to AI requirements.
The data abstraction capabilities provided in TELOS add to the already powerful data type exten sion capabilities of PASCAL. The user defines a problem-specific data type by including details of representation and implementation within a definitional scoping called a "capsule". The pro cedures and functions which realize possible prim itive operations on objects of the type being de fined are an integral part of the definition, and the objects are characterized and used in terms of these defining operations. Features are provided for specifying and controlling how objects of a capsule-defined type are to be fitted into the TELOS associative data base.
An AI programmer might use TELOS capsules to define data types ranging from low-level lists (of the kind built into most other AI languages) to theoretically significant data structures. Such high-level abstract type definitions might imple ment, for example, case-slotted prototypes (frames), concept-dependency graphs, or production "conditionals". By emphasizing abstraction mecha nisms rather than high-level constructs (e.g., frames as a basic data type, as in KRL), it has been possible to minimize theoretical bias in the language. Consequently TELOS should be usable for investigation of numerous competing theories, for example, in the area of knowledge representation.
The control abstraction capabilities provided 1n TELOS enable convenient user definition of the novel kinds of control regimes which are investi gated in AI research, that is, those which realize alternative problem-solving strategies. Just as TELOS capsules localize data represe~tation de tails, TELOS "overseers" localize 1nterprocess control-transfer and communication details needed to realize a desired control regime.
An AI programmer might use TELOS overseers to d~fine control regimes ranging from blind, depth flrst backtracking (of the kind built into many recent AI languages) to quasi-parallel processing to sophisticated planning strategies. Features are provided which enable the procedures and proc esses controlled by an overseer to be identified and invoked descriptively as well as by name.
Besides its abstraction mechanisms, TELOS contains other facilities which can contribute to t~e buildi~g of well-structured programs, starting w1th the r1ch set of program structuring facili ties already available in PASCAL. If used cor rectly, these facilities can result in modular ized, hierarchical programs with reasonably com prehensible control and data flows.
Many language capabilities very useful in AI programming directly conflict with principles of good program structuring. The TELOS design at tempts to mediate the conflicts with compromises which give AI programmers some of the advantages of both worlds. Examplesl (1) TELOS allows the construction of a heterarchical control regime but centralizes into one location its communication and control-transfer activities. (2) TELOS allows specification of conditional interrupts (i.e., de mons) but subsumes them under a general, unified event handling scheme. The scheme enables signal, escape (error), and suspend events to be explicit ly declared and activated, and (whether explicit or not) to be operative only within certain clear ly circumscribed and easily determined scopes. (3) TELOS allows creation of alternative data base versions (i.e., contexts) but requires that switching of the current execution context be ac complished with the structured forml
InContext <context> Do <statement>
Several benefits should accrue for AI from such a language designl
- AI programs are highly complex. A language like TELOS provides many aids lo managing and con taining program complexity.
- Improved AI programmer productivity, as is possible with the features TELOS provides, will mean improved AI research productivity.
- Much of AI programming involves putting preliminary ideas and hypotheses into programs and then changing the pro~rams as suggested by experi ence with them. The 1mproved program comprehen sibility possible with TELOS can make this kind of evolutionary programming much easier and more ef ficient.
- One of the motivations of work on AI is de velopment of a general theory of intelligence. Such theory development will be expedited if the program abstractions representing high-level theo retical constructs are as distinct and as well-defined as they can be in TELOS programs.
@inproceedings{ijcai1977p215,
title = {TELOS, A Language for Building Well-Structured AI Models},
author = {Travis, Larry and Honda, Masahiro and LeBlanc, Richard and Zeigler, Stephen F.},
booktitle = {Proceedings of the Fifth International Joint Conference on
Artificial Intelligence, {IJCAI-77}},
publisher = {International Joint Conferences on Artificial Intelligence},
editor = {Raj Reddy},
pages = {917---1},
year = {1977},
note = {Languages and Systems},
url = {https://ijcai.org/proceedings/1977/215},
}
TY - CPAPER TI - TELOS, A Language for Building Well-Structured AI Models AU - Travis, Larry AU - Honda, Masahiro AU - LeBlanc, Richard AU - Zeigler, Stephen F. T2 - Proceedings of the Fifth International Joint Conference on Artificial Intelligence, IJCAI-77 PB - International Joint Conferences on Artificial Intelligence PY - 1977 SP - 917 EP - -1 UR - https://ijcai.org/proceedings/1977/215 ER -Copied
Machine-readable versions of this paper, for tools and AI agents:
Welcome
Future Conferences
AI Journal Division
IJCAI Proceedings
About IJCAI
IJCAI Awards