Mark Humphrys

Bayesian Imitation of Human Behavior in Interactive Computer Games

Modern interactive computer games provide the ability to objectively record complex human behavior, offering a variety of interesting challenges to the pattern recognition community. Such recordings often represent a multiplexing of long-term strategy, mid-term tactics and short-term reactions, in addition to the more low-level details of the players movements. In this paper, we describe our work in the field of imitation learning; more specifically, we present a mature, Bayesian-based approach to the extraction of both the strategic behavior and movement patterns of a human player, and their use in realizing a cloned artificial agent. We then describe a set of experiments demonstrating the effectiveness of our model

How My Program Passed the Turing Test

In 1989, the author put an ELIZA-like chatbot on the Internet. The conversations this program had can be seen – depending on how one defines the rules (and how seriously one takes the idea of the test itself) – as a passing of the Turing Test. This is the first time this event has been properly written. This chatbot succeeded due to profanity, relentless aggression, prurient queries about the user, and implying that they were a liar when they responded. The element of surprise was also crucial. Most chatbots exist in an environment where people expectto find some bots among the humans. Not this one. What was also novel was the onlineelement. This was certainly one of the first AI programs online. It seems to have been the first (a) AI real-time chat program, which (b) had the element of surprise, and (c) was on the Internet. We conclude with some speculation that the future of all of AI is on the Internet, and a description of the “World- Wide-Mind” project that aims to bring this about.

Building a Hybrid Society of Mind Using Components from Ten Different Authors

Building large complex minds is difficult because we do not understand what the necessary components are or how they should interact. Even if the components were known it is difficult to see a situation where a single lab would have all the necessary expertise and manpower to be able to build the mind. One possible answer is to distribute the various components of the mind throughout the Internet, by letting different groups or individuals build parts of the mind. What would be needed then is a protocol that can be used to allow the components to interact, and a mechanism for managing the independent failure of components, and the switching between the required parts of the mind. This paper describes the first attempt to build such a mind. Using components that were developed independently by ten different authors, we have built a mind to solve a specific problem. Although the problem is a simple blocks world implementation, the issues dealt with, as well as the technology used, are relevant for larger and more complex minds. Robust minds require duplication and distribution. The Internet provides the ideal way to build in just such features into artificial minds.

A Framework for Scaling Up Distributed Minds

An important topic in artificial intelligence is the modular decomposition of problems and solutions into smaller ones. Modular, “hybrid” solutions tend to be built from components mostly created by the same programmer or the same team. What if we already have many solutions online - programs created by a number of different authors and hosted as services - and wish to compose them into a larger, distributed hybrid program which performs better than the individual components? Can this be done automatically? The World-Wide Mind (W2M) project attempts to scale up artificial intelligence by distributing action-selecting agents (which we call “minds”) and problem environments (which we call “worlds”) as services on the internet, and by allowing minds to call other minds and thus facilitate building hybrid minds from many programs which may have been written by many authors. This paper gives a general overview of the W2M architecture and ongoing work examining the possibility of automatically constructing hybrid minds.

A novel application of Web Services in Computer Science education

Many computer based tools exist which facilitate collaboration among groups of students. The majority of such tools, however, simply serve as communication devices in substitute for face to face communication. This paper describes how computer science students can use their skills to collaborate in a new way. Using Web services to develop interactive content, a group of undergraduate computer science students collaborated to develop solutions for an assessment in an artificial intelligence course. The term Web services describes a new and exciting family of technologies for connecting online programs. Although they are rapidly becoming popular in industrial applications, we feel they have been ignored in educational environments, where they can serve as simple mechanisms for collaborating to develop content. This type of collaboration requires technical skill of a level higher than that of a computer user, but it provides collaborating computer science students with an excellent opportunity to employ the networked computer as an enhanced learning tool, and not simply a communication device

An experimental system for real-time interaction between humans and hybrid AI agents

There is an emerging belief among AI researchers that intelligence is the product of specialised subsystems collaborating in a type of network of the mind. Several prototype systems have been developed as part of the World-Wide-Mind project [6] that facilitate the on-line deployment of hierarchically structured, multi-author, AI agents that we call minds. These agents function in on-line user defined environments that we call worlds. In this paper we describe the latest prototype system developed as part of the World-Wide-Mind project, the XAI (Experimental Artificial Intelligence) Server. Unlike its predecessors the XAI Server is a multi-agent system (MAS) allowing minds to run concurrently in the same on-line environment, facilitating inter-mind communication, cooperation and competition. The XAI Server also adds real-time interaction between humans and AI agents. Along with the potential for gauging the effectiveness of AI algorithms this raises issues relating to timing and knowledge representation that previous prototypes had no need to address. The XAI Server is loosely coupled with a scheduled threading model that makes it far more scalable than its predecessors. Capable of communicating with a standard Web browser the XAI Server moves the responsibility for content generation from the server to the client via a proprietary user interface language. We demonstrate how this system works with a sample ChatWorld and four example minds. Finally we discuss our future plans for enhancing the system to facilitate 2D and 3D world authoring.

Problems with Internet and Library Usage for Secondary School Children

This research consisted of investigating seven hypotheses using the following components: 1. Observation of forty-three secondary school children using the Internet and the library to complete five tasks. 2. An interview was held with all the participants in the study that was audio taped and subsequently transcribed. 3. An on-line form was constructed so that the students could nominate their favourite educational web sites for each class subject. The participants were given five tasks and had to find the answers using two of the most common forms of information retrieval found in Irish schools, the Internet and the school library. Subsequently they were asked twenty questions about their opinions on aspects of the library and the Internet. Points of interest here included that the majority of participants felt that the Internet is faster, easier to use, and better overall than the library, even though it was proven not to be the case. It was also found that the participants nominated sites by domain name without actually investigating if the domain name had any reference to the subject in question. The observation, interview and voting data was then analysed using SPSS to investigate the seven hypotheses. These findings were then reported, discussed and ideas for future study were recommended. Proposed technical and teaching solutions to problems uncovered in this research are also outlined. The above findings have implications for search engine design, the curriculum of the Irish education system, and for teachers in how they use both the Internet and the library to their full potential.