Dr. Mark HumphrysBSc. Joint Hons (UCD), PhD (Cambridge)
Lecturer,
The World-Wide-Mind project:
w2mind.org
|
 Photo 2002. |
Teaching
I lecture various Computer Science courses here in DCU. See my Teaching page, or jump direct to the course pages for:
Undergrads looking for 3rd or 4th year projects can see my list of Undergrad project ideas.
Postgrads on the MSc in Bioinformatics can see my list of Masters project ideas.
Admin
I am currently the Working Papers Librarian.
Research
I am interested in non-symbolic (or sub-symbolic) Artificial Intelligence. I am interested in the origins of intelligence, both the long evolutionary history of the species, and the long developmental history of each individual. I am interested in the vast substrate of animal sensorimotor skills and sub-linguistic knowledge representation that lies beneath all the high-level (and recently-evolved) human cognitive skills that we focus so much attention on. This substrate to me is where the hard part of AI lies. You can start at my Research page, or go direct to my Publications list.
Action Selection
In particular, my research is on sub-symbolic decision-making, or Action Selection among competing, co-operating and overlapping non-symbolic behaviours. I am interested in highly-decentralised Society of Mind models.
For a simple overview,
there are some
Introductory Movies
that you can play.
Multiple Minds in the same Body: My PhD thesis: "Action Selection methods using Reinforcement Learning" (1997) introduces "W-learning" - a Model of Mind whereby different parts of the mind modify their behaviour based on whether or not they are succeeding in getting the body to execute their actions. Where this is headed is towards a complex, overlapping, competing, sub-symbolic Society of Mind based on Reinforcement Learning.
I think I may have been the first to use Unhappiness based models of Action Selection using Reinforcement Learning numbers - that is, based on differences between the Q-values.
In terms of the World-Wide-Mind (see below), my PhD can now be seen as a model of AI where parts of the mind do not understand each other (e.g. could be written by different authors). My PhD can also now be seen as a model of AI that can survive broken links.
My PhD "family tree" (Who supervised who)
The World-Wide-Mind
I have a new idea for helping AI scale up, and enabling the construction of large, complex minds by teams of multiple dispersed authors. This idea is called the "World-Wide-Mind". I have started a "World-Wide-Mind" research group and we have set up a portal site w2mind.org.For an introduction to this idea see: Humphrys, Mark (2001), "The World-Wide-Mind: Draft Proposal", Dublin City University, School of Computing, Technical Report no. CA-0301, February 2001.
For short introductions see the papers at ECAL-01 or SAB-02.
Updates to the project are at the
World-Wide-Mind blog.
The Turing Test
I have some writings on The Turing Test because of continued interest in "MGonz", an AI chatbot I hooked up to the Internet in 1989. This story is told in "How my program passed the Turing Test". This program is now online again, and you can talk to it once more.My program was certainly one of the first AI programs online. I think it was the first (a) AI real-time chat program, which (b) had the element of surprise, and (c) was on the Internet.
MGonz is finally written up as a book chapter:
Humphrys, Mark (2008),
"How my program passed the Turing Test",
Chapter 15 of
Parsing the Turing Test: Philosophical and Methodological
Issues in the Quest for the Thinking Computer,
Robert Epstein, Gary Roberts and Grace Beber (eds.),
Springer, 2008.
AI in general
I have also some writings on AI in general. I have written for New Scientist a Popular-science introduction to the biologically-inspired type of AI. See also my Philosophy and Future of AI page.I gave a talk on the future of AI, "The Hardest Problem in the History of Science", at the Institute of Contemporary Arts, London, Feb 2000.
Computers and Internet
I have been a heavy user of the Internet since 1987, and have enjoyed watching it grow up over those years, driven almost entirely by the brilliant idea of embedding the navigation in hypertext. I think I have learnt a few things over this time, and there are some slightly less formal writings and links on my Computers and Internet page.I have put up a few bits and pieces on The Internet in the 1980s.
My website has been running since 1994. You can go direct to an article I wrote in 1999 for the Irish Times, which summarises a lot of my experience of the Internet over the years, called "Why on earth would I link to you?".
Programs
Some programs I have written for my website:
- A custom CGI script
search engine
(for my sub-site of the CA website only).
- My sub-site functions as an error-tolerant web server. When you enter a bad URL, instead of giving you a bland "404 Not Found" message, the web server redirects to my own Error Handling script, which tries to guess the URL you should have typed, using case-insensitive and partial-string matching. Try it out by entering a bad address (within my sub-site only).
Programs I have written for my email:
- My scheme for stopping spam.
|
Action Selection
|
