Nils J. Nilsson’s research while affiliated with Stanford University and other places
What is this page?
This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.
Machine Intelligence 14 contains material presented at the Anglo-Janpanese workshop of Novemver 1993 held at the Hitachi Research Laboratory. It marks the 70th birthday of Donald Michie, the founder of the series. The contents is divided into the following subjects: complex decision taking, inductive logic programming, applied machine learning, dynamic control, and computational learning theory. Applications include controlling a steel mill, discovery of protein structural constraints, and qualitative control for dynamic systems.
In its early stages, the field of artificial intelligence (AI) had as its main goal the invention of computer programs having the general problem solving abilities of humans. Along the way, there developed a major shift of emphasis from general-purpose programs toward "performance programs"---ones whose competence was highly specialized and limited to particular areas of expertise. In this paper I claim that AI is now at the beginning of another transition---one that will re-invigorate efforts to build programs of general, human-like competence. These programs will use specialized performance programs as tools, much like humans do. Keywords: autonomous agents, general problem solving, habile systems Copyright c fl1995 Nils J. Nilsson [This paper is being submitted to the AI Magazine.] 1 Diversions from the Main Goal Over forty years ago, soon after the birth of electronic computers, people began to think that human levels of intelligence might someday be realized in computer program...
This major collection of short essays reviews the scope and progress of research in artificial intelligence over the past two decades. Seminal and most-cited papers from the journal Artificial Intelligence are revisited by the authors who describe how their research has been developed, both by themselves and by others, since the journals first publication. The twenty-eight papers span a wide variety of domains, including truth maintainance systems and qualitative process theory, chemical structure analysis, diagnosis of faulty circuits, and understanding visual scenes; they also span a broad range of methodologies, from AI's mathematical foundations to systems architecture. The volume is dedicated to Allen Newell and concludes with a section of fourteen essays devoted to a retrospective on the strength and vision of his work.
Sections/Contributors Artificial Intelligence in Perspective, D. G. Bobrow • Foundations, J. McCarthy, R. C. Moore, A. Newell, N. J. Nilsson, J. Gordon and E. H. Shortliffe, J. Pearl, A. K. Mackworth and E. C. Freuder, J. de Kleer • Vision, H. G. Barrow and J. M. Tenenbaum, B. K. P. Horn and B. Schunck, K. Ikeuchi, T. Kanade • Qualitative Reasoning, J. de Kleer, K. D. Forbus, B. J. Kuipers, Y. Iwasake and H. A Simon • Diagnosis, R. Davis, M. R. Genesereth, P. Szolovits and S. G. Pauker, R. Davis, B. G. Buchanan and E. H. Shortliffe, W. J. Clancey • Architectures, J. S. Aikins, B. Hayes-Roth, M. J. Stefik et al • Systems, R. E. Fikes and N. J. Nilsson, E. A Feigenbaum and B. G. Buchanan, J. McDermott. Allen Newell. H. A. Simon, M. J. Stefik and S. W. Smoliar, M. A. Arbib, D. C. Dennett, Purves, R. C. Schank and M. Y. Jona, P. S. Rosenbloom and J. E. Laird, P. E. Agre.
Bradford Books imprint
A formalism is presented for computing and organizing actions for autonomous agents in dynamic environments. We introduce the notion of teleo-reactive (T-R) programs whose execution entails the construction of circuitry for the continuous computation of the parameters and conditions on which agent action is based. In addition to continuous feedback, T-R programs support parameter binding and recursion. A primary difference between T-R programs and many other circuit-based systems is that the circuitry of T-R programs is more compact; it is constructed at run time and thus does not have to anticipate all the contingencies that might arise over all possible runs. In addition, T-R programs are intuitive and easy to write and are written in a form that is compatible with automatic planning and learning methods. We briefly describe some experimental applications of T-R programs in the control of simulated and actual mobile robots. Comment: See http://www.jair.org/ for any accompanying files
New ideas are presented for computing and organizing actions for autonomous agents in dynamic environments-environments in which the agent's current situation cannot always be accurately discerned and in which the effects of actions cannot always be reliably predicted. The notion of 'circuit semantics' for programs based on 'teleo-reactive trees' is introduced. Program execution builds a combinational circuit which receives sensory inputs and controls actions. These formalisms embody a high degree of inherent conditionality and thus yield programs that are suitably reactive to their environments. At the same time, the actions computed by the programs are guided by the overall goals of the agent. The paper also speculates about how programs using these ideas could be automatically generated by artificial intelligence planning systems and adapted by learning methods.
The theoretical foundations of the logical approach to artificial intelligence are presented. Logical languages are widely used for expressing the declarative knowledge needed in artificial intelligence systems. Symbolic logic also provides a clear semantics for knowledge representation languages and a methodology for analyzing and comparing deductive inference techniques. Several observations gained from experience with the approach are discussed. Finally, we confront some challenging problems for artificial intelligence and describe what is being done in an attempt to solve them.
... A physical version of his abstract world, Shakey's "real environment," was constructed for him within SRI's facilities. Resembling a poorly designed open office layout with linoleum floors, low walls, and an acoustical ceiling with fluorescent lighting laid out to minimize shadows, the space was designed for Fig. 9 A diagram of Shakey and his components (Nilsson 1984;Nilsson et al. 1969;Raphael et al. 1971) the benefit of Shakey's vision system, with paint colors and other high-contrast finishes (especially the dark baseboard) to help his fledgling image processing and edge-detection algorithms. The objects that Shakey was tasked with navigating around, moving, and organizing were rectangular or wedge-shaped and painted in alternating specular paint colors on each face, also to emphasize their edges. ...
... Single-Agent Approaches. Recent advances in large language models (LLMs) such as GPT-4 [33] have renewed interest in the development of autonomous agents that can solve tasks on behalf of people [32,59,16,60,65,49,74,43]. These modern agents have shown remarkable skills in software development [55,76,66,63], web manipulation [8,75,79,31,1], manipulation of general graphical user interfaces [73,61,3,34], and other domains [37,54]. ...
... The above algorithm is simple to apply, but it also suffers from the disadvantages of many extended nodes and low search efficiency. The emergence of the A* algorithm effectively solves this problem, and the A* algorithm was first proposed by Peter Hart in 1968 based on Dijkstra's algorithm 12 . It retains the advantages of Dijkstra's algorithm and reduces the number of expansion nodes. ...
... Macro-actions are a form of temporal abstraction to tackle long-horizon planning, by reducing the planning depth linearly and thus the planning complexity exponentially. For deterministic planning and MDPs, a macro-action can be represented as a sequence of primitive actions or a local trajectory [6,10,16,37]. Such trajectories can be extracted from past experience, typically generated by planning using primitive actions. ...
... Our goal in this study is to better understand how the knowledge needed for rapid action-sequence planning might be stored and processed in the brain. To this end, we develop a spiking-neuron model that plans action sequences by chaining together action preconditions and effects (Fikes & Nilsson, 1971 ) while interacting with a simulated environment . Each planning step selects from actions that are related to available objects, in order to constrain each decision and allow planning to proceed quickly (about 100ms of simulated time per step). ...
... Based on this framework, we propose two novel Banditbased MCTS algorithm designed for heuristic search including Classical Planning (Fikes et al., 1972). Our novel bandits, UCB1-Power and UCB1-Uniform, each models a special case of GP distribution to avoid the numerical difficulty of estimating its parameters. ...
... With an AI that does not need to take a rest and sleep, the probabilities of error are almost nil and that leads to perfection and faster production. In other words, AI will minimize the cost of living because it reduces the need for human manpower, thus reducing operational costs [16,18]. ...
... These techniques are the simplest form of artificial intelligence and mimic the reasoning of a human expert in solving a knowledge-intensive problem. In other words, AI-RBSs encode human expert knowledge about a specific topic into an automated system; they are often used to comprise an expert system [43,44]. An AI-RBS reproduces deductive reasoning mechanisms by employing logic rules made of conjunctions of conditions to verify a set of actions to execute [45]. ...
... Unfortunately, to the best of our knowledge, all the representation frameworks for ontologies which are rooted in probability theory exhibit lossy reasoning or have counterintuitive restrictions on their flexibility. The probabilistic DLs based on Nilsson's probabilistic logic [4] experience decay in relative precision during reasoning due to their expression of probabilities as intervals. Approaches using Bayesian Networks (BNs) [5], such as BayesOWL [6], MEBN/ PR-OWL [7], and P-CLASSIC [8], contain a representation granularity mismatch: Bayesian Networks require complete specification of the domain's probability distribution with no incompleteness, but ontologies have a finer granularity which allows for incompleteness. ...
... Algebra has had a significant impact on the field of AI, particularly in the areas of queries and data representation [25]. It has proven to be instrumental in enhancing the efficiency and effectiveness of these processes [26]. ...
