PITTSBURGH (AP) — Researchers are trying to plant adigital seed for artificial intelligence by
letting a massive computersystem browse millions of pictures and decide for itself what they allmean.The
system at Carnegie Mellon University is called NEIL,short for Never Ending Image Learning. In mid-July, it
began searchingthe Internet for images 24/7 and, in tiny steps, is deciding for itselfhow those images
relate to each other. The goal is to recreate what wecall common sense — the ability to learn things without
beingspecifically taught.It’s a new approach in the quest to solvecomputing’s Holy Grail: getting a machine
to think on its own using aform of common sense. The project is being funded by Google and theDepartment of
Defense’s Office of Naval Research."Any intelligentbeing needs to have common sense to make
decisions," said AbhinavGupta, a professor in the Carnegie Mellon Robotics Institute.NEILuses advances
in computer vision to analyze and identify the shapes andcolors in pictures, but it is also slowly
discovering connectionsbetween objects on its own. For example, the computers have figured outthat zebras
tend to be found in savannahs and that tigers look somewhatlike zebras.In just over four months, the network
of 200processors has identified 1,500 objects and 1,200 scenes and hasconnected the dots to make 2,500
associations.Some of NEIL’scomputer-generated associations are wrong, such as "rhino can be a kindof
antelope," while some are odd, such as "actor can be found in jailcell" or "news anchor
can look similar to Barack Obama."But Guptasaid having a computer make its own associations is an
entirelydifferent type of challenge than programing a supercomputer to do onething very well, or fast. For
example, in 1985, Carnegie Mellonresearchers programed a computer to play chess; 12 years later, acomputer
beat world chess champion Garry Kasparov in a match.CatherineHavasi, an artificial intelligence expert at
the MassachusettsInstitute of Technology, said humans constantly make decisions using"this huge body of
unspoken assumptions," while computers don’t. Shesaid humans can also quickly respond to some questions
that would take acomputer longer to figure out."Could a giraffe fit in your car?"she asked.
"We’d have an answer, even though we haven’t thought aboutit" in the sense of calculating the
giraffe’s body mass.RobertSloan, an expert on artificial intelligence and head of the Departmentof Computer
Science at the University of Illinois, Chicago, said theNEIL approach could yield interesting results
because just usinglanguage to teach a computer "has all sorts of problems unto itself.""WhatI
would be especially impressed by is if they can consistently say’zebra, zebra, zebra’ if they see the animal
in different locations,"Sloan said of the computers.Gupta is pleased with the initialprogress. In the
future, NEIL will analyze vast numbers of YouTubevideos to look for connections between objects."When
we started the project, we would not sure it would work," he said. "This is just the
start."NeitherMountain View, Calif.-based Google Inc. nor the Office of NavalResearch responded to
questions about why they’re funding NEIL, butthere are some hints. The Naval Research website notes that
"today’sbattlespace environment is much more complex than in the past" and that"the rate at
which data is arriving into the decision-making system isgrowing, while the number of humans available to
convert the data toactionable intelligence is decreasing."In other words, computersmay make some of the
decisions in future wars. The Navy’s website notes:"In many operational scenarios, the human presence
is not an option."NEIL’smotto is "I Crawl, I See, I Learn," and the researchers hope to
keepNEIL running forever. That means the computer might get a lot smarter.Or it might
not.___Online:www.neil-kb.com___Follow Kevin Begos at https://twitter.com/kbegosCopyright 2013 The
Associated Press. All rightsreserved. This material may not be published, broadcast, rewritten
orredistributed.