artificial intelligence

Navigation

computing / technology: artificial intelligence

Overview

The endeavor to create artificial intelligence (AI) via electronic computational methods, which is generally agreed to have begun in the 1950s, has turned out to be considerably more difficult than originally imagined.

Partly as a result of this, the field has split into areas whose goals vary in their reach:

• narrow AI generally refers to carefully-written programs which seem to respond with a certain amount of intelligence in certain narrowly-defined areas; "programs that solve particular, highly specialized types of problems" (2007-10-18 KA).
• general AI refers to what we more commonly think of as "intelligence", i.e. the ability to solve new types of problems; adaptability, flexibility; "programs with the autonomy and self-understandings to come to grips with novel problem domains and hence solve a wide variety of problem types" (2007-10-18 KA).

Early attempts suffered greatly from lack of computational power; Moore's Law (which more or less doubles available CPU power in several measurable ways every 1.5 years or so) has now vastly increased the available power, and the best estimates are that an average desktop PC will possess approximately the computing power of a human brain sometime before 2030. What remains is to create the software to allow it to think like one.

Efforts have been focused in a number of areas:

• scripted conversational strategies, e.g. AIML
• "common sense" engines, mainly Cyc
• artificial neural networks (these seem to have fallen out of favor, but nonetheless showed great promise as a tool)

Links

Projects

• A.L.I.C.E. Artificial Intelligence Foundation: development of scripted conversational AI using Artificial Intelligence Markup Language (AIML): shows potential for adaptivity if combined with other AI tools. The main problem with allowing the engine to "learn" from interactive users is the "trust" issue; this will probably require the development of a social skills engine, i.e. keeping track of who tells the AI what, correlating individuals with corroboration of facts received, and deciding when there is enough corroboration to accept received information as "fact" (and how to do so, i.e. how should the AI remember that this is a "new" fact, with perhaps some indication of where it came from?)
• Artificial General Intelligence Research Institute (AGIRI) (wiki)
• Cognitive Code Corporation's SILVIA technology sounds suspiciously like a proprietary implementation of an AIML engine...
• The "Dexter" robot:
  • 2007-03-01 Dexter Jumps
  • 2007-02-26 Dexter Walks
• 2007-06-01 CB2 robot:
  • Giant robot baby helps team study childhood
  • CB2 baby humanoid robot (includes pictures and video) Team hopes to simulate a 3-year-old within the next 4 years
  • see also wikipedia:Uncanny Valley (robotics phenomenon)
• Cyc: open-source project to develop "common sense" software

Articles

• 2007-10-18 KA AI Meets the Metaverse: Teachable AI Agents Living in Virtual Worlds by Ben Goertzel

News

• 2008-03-03: Out of the Blue: "The Blue Brain project is now at a crucial juncture. The first phase of the project—"the feasibility phase"—is coming to a close. The skeptics, for the most part, have been proven wrong. It took less than two years for the Blue Brain supercomputer to accurately simulate a neocortical column, which is a tiny slice of brain containing approximately 10,000 neurons, with about 30 million synaptic connections between them. "The column has been built and it runs," Markram says. "Now we just have to scale it up." Blue Brain scientists are confident that, at some point in the next few years, they will be able to start simulating an entire brain." They are starting by closely modeling structures in a rat brain -- and checking the results against actual rat brain responses. (This does not appear to be the same group as the 1/2-mouse-brain simulation reported on 2007-04-27.)
• 2008-03-01: Japanese robots enter daily life
• 2007-04-27: (1/2) Mouse brain simulated on computer
• 2006-09-14: Experimental AI Powers Robot Army and more
• found 2006-05-15: Artificial Intelligence Algorithms For Top-Down AI at SourceForge
• 2006-02-06 Software That Learns by Doing (also has implications for the Intelligent Design controversy)
• 2006-01-14 Crouching Tiger, Hidden Robot
• 2006-01-12 St. Lawrence of Google: Google is working on a massive global computing grid; wants to "build the machine that will pass the Turing test"
• 2005-11-11 Firm to introduce robot lawyers

Video

• Boss NQE Test, Sunday, Oct. 28 - Darpa Urban Challenge: self-driving car
  • Commentary: Pharyngula (2008-04-12)