Behind Artificial Intelligence, a Squadron of Bright Real People--

"Behind Artificial Intelligence, a Squadron of Bright Real People"
New York Times (10/14/05); Markoff, John

That five robots completed the 132-mile course through the Nevada desert last weekend constituted a major victory for artificial intelligence, a field that has long been beset by disappointment and the failure to live up to its expectations. Guided only by GPS tracking, the winning entry came from a team of Stanford researchers whose modified Volkswagen dubbed Stanley completed the course in six hours and 53 minutes. The move from logic and rule-driven systems to an approach more concerned with probability and statistics has helped researchers advance artificial intelligence beyond mundane tasks of voice recognition and into more complex applications such as driving. Thanks to the incorporation of sensors in new artificial intelligence systems, devices such as Stanley can make decisions based on probability. The Stanford lab created the first autonomous vehicle in 1975, which was moving in increments of two feet by the end of the decade, though speculative interest cooled considerably as development stagnated in the 1980s. Automobile safety has generated renewed interest in artificial intelligence, particularly from government agencies such as Darpa, which sponsored the Grand Challenge in Las Vegas. While some are concerned that Darpa's funding of artificial intelligence will relegate the field to military contractors working on classified projects, ACM President David Patterson reminds us that the Stanford victory demonstrates the ability of the academic community to achieve great success on a limited budget. "This is consistent with the history of our field," said Patterson. "This demonstrates the importance of the participation of government-funded academics."
Click Here to View Full Article

"The Mind of an Inventor"

Newsweek (10/10/05) Vol. 146, No. 15, P. 48; Levy, Steven

Developing futuristic but practical technologies is the motivation behind Applied Minds, a company whose gadgets are a testament to the childlike imagination of its co-founder, inventor Danny Hillis. Applied Minds' first commercialized product, developed in collaboration with the Herman Miller office furniture company, was Babble, a device designed to increase the privacy of office cubicle workers by masking their conversation with a soundtrack of scrambled, nonsensical vocal fragments. Meanwhile, Hillis' company is developing a personal robot dubbed "the mule" with Northrop Grumman that seeks to relieve soldiers from the burden of carrying their own water, communications equipment, and batteries. The mule, which trails behind the soldier, produces water from air and has broadband built-in. Another fascinating Applied Minds product is the 2.5-D Display, a table that can present topographical information of any location on Earth as a detailed physical model. As a student at MIT, Hillis envisioned the concept of a thinking machine, which inspired him to re-imagine the computer's "brain" as a combination of thousands of interoperating processors rather than one processor. His idea of "parallel processing" inspired a doctoral thesis as well as a company, Thinking Machines. Despite his many accolades and patents, Hillis asserts that "people tend to overestimate the individual inventor and underestimate the system that makes their inventions real."

Click Here to View Full Article

Sony Researchers Create 'Curious' Aibos

"Sony Researchers Create 'Curious' Aibos"

IDG News Service (06/14/05); Kallender, Paul

Researchers at the Sony Computer Science Laboratory in Paris have announced the possible creation of synthetic curiosity in the form of the Aibo ERS-7 robot dog. Advances in computing power have led to systems that appear to have the ability to react based on pre-programmed, task-defined algorithms, but these devices can only act within a given set of parameters. The Sony researchers, however, have equipped Aibo with a "metabrain" algorithm that continually forces a learning algorithm to look for new and more challenging tasks and to abandon tasks that do not appear to lead anywhere, in essence creating virtual boredom, curiosity, and a desire to learn. In repeated experiments, various Aibos, after being placed in play pens with balls, learned progressively to swivel their legs and heads, wiggle, crawl, and crawl and hit and follow a ball placed in front of them despite not being programmed to do so. "What we have done is give Aibo an ability to have its learning defined by the quality of learning," explains researcher Pierre-Yves Oudeyer. "So it's learning to classify its sensory space and progressively structure that space." The researchers admit that giving Aibo the cognitive intelligence comparable of even a two-month-old human infant will be difficult.

Click Here to View Full Article

Understanding Agent System

Understanding Agent Systems

Mark d'Inverno and Michael Luck

Series: Springer Series on Agent Technology,

2nd rev. and extended ed., 2004, XVIII, 240 p., Hardcover

ISBN: 3-540-40700-6

An introductory level text on agent system. Authers tried to define a common set of terminology for this research area.

Topics:

• The Agent Landscape


• The SMART Agent Frameworks


• Agent Relationships


• An Operational Analysis of Agent Relationships


• Sociological Agents


• The Contract Net as a Goal Directed System


• Computational Architecture for BDI Agents


• Evaluating Social Dependence Networks


• Z Specification Language

[Recommand]**

[Amazon]*****

Please see
Springer for more info.

Reasoning about Rational Agents

Reasoning about Rational Agents

Wooldridge, Michael, J.

2000, MIT. ISBN: 0-262-23213-8

Aside of the strong theories like Decision Theory or Game Theory (in muti-agent senario), which give few clues to implementation, the author proposed a logic based practical reasoning decision making from the "belief-desire-intention" (BDI) model.
The author also introduced a formal framework, LORA, (Logic Of Rational Agents), as a tool for understanding and implementing the BDI.

[Recommand]***
See MIT Press for more.

Artifitial Intelligence: A Modern Approach

Artifitial Intelligence: A Modern Approach(2nd Ed.) is a textbook for the first class in AI. Since I finally decided to devote myself into this field, I will read this BIG book on my own.

Also, Learn Prolog Now! by Patrick Blackburn, Johan Bos and Kristina and
Striegnitz, is also downloadable via the link.

This blog will then become my notes for the first step study of AI, Prolog, and first order logic.

Go!