NASA satellites to use artificial intelligence

greenspun.com : LUSENET : Unk's Wild Wild West : One Thread

http://www.cnn.com/2001/TECH/space/06/04/casper/index.html

NASA satellites to use artificial intelligence

June 4, 2001 Posted: 3:46 PM EDT (1946 GMT)

By Richard Stenger

CNN

(CNN) -- Software that can make decisions without guidance from humans will guide a constellation of spacecraft that launch in 2002, according to NASA.

The Artificial Intelligence program will direct three sibling satellites, allowing them to respond to events on their own, said NASA's Jet Propulsion Laboratory, which will manage the mission.

The miniature satellites, each of which weighs less than 33 pounds (15 kg), will fly in a tight formation in orbit as part of the Three Corner Sat mission.

"The onboard software will perform decision-making function for the spacecraft," said lead project scientist Steve Chien. "Like a brain that uses inputs from the eyes and ears to make decisions, this software uses data from spacecraft sensors, such as cameras, to make decisions on how to carry out the mission."

NASA has dabbled before in using artificial intelligence to govern a spacecraft. During one experiment AI software controlled the Deep Space I robot ship, currently en route to an encounter with a comet.

But the latest AI program will direct the mission without interruption for at least three months, according to JPL, which is based in Pasadena, California.

Conventional satellites engage in a laborious, protracted form of communications with ground controllers. All science data is beamed back to Earth, good or bad.

But the AI software, known as Continuous Activity Scheduling, Planning Execution and Replanning (CASPER), will have the ability to make real-time decisions based on the images it acquires and send back only those that it considers important.

Moreover, when a satellite faces hazards in space, perhaps an intense solar storm or unexpected blurry vision, mission engineers must come up with new computer instructions from scratch and send them to the probe.

But the new software will enable spacecraft to attempt to come up with their own solutions, rather than rely on time-consuming assistance from Earth.

"This capability represents a significant advance from traditional ground-based operations and offers promise to dramatically increase mission science for this and future missions," said Colette Wilklow, a project researcher and engineering student at the University of Colorado.

The mission, a joint project of NASA, universities and the military, will showcase satellite advances in stereo imaging, formation flying and autonomous command capability, JPL said.

-- (in@the.news), June 05, 2001

Answers

Maybe someone more knowledgeable than me will comment----is there truly such a thing as AI? Aren't the "decisions" that the satellites will make still determined by S/W instructions, albeit complex?

I don't agree that a nonliving entity is capable of "intelligence". That includes IBM's chess champion Deep Blue. If DB continues to defeat human chess masters, I think that says more about the nature of chess than it says about intelligence.

Can AI write a poem, make a joke, do anything independent, spontaneous, original, creative?

-- Lars (larsguy@yahoo.com), June 05, 2001.


"is there truly such a thing as AI?"

Certainly not high order intelligence of the sort demonstrated by humans. Or pigs, dogs or parrots for that matter. At the moment, AI has acheived a level of independent intelligence that could better be compared to a cockroach or other insects. So long as the range of input does not exceed the expected range, AI programs can process a limited subset of the input variables and form a roughly appropriate response. As soon as the unexpected arises, they tend to fail rapidly. Same witrh insects. That is why so many hatch and so few live. In space the range of input is likely to be pretty narrow and pose a small set of problems to solve giving AI a nice, clean environment in which to operate.

I agree with you about Deep Blue. Chess is a bounded system with a very limited set of rules. The number of legal chess moves for any configuration of the board is calculable. That is what makes Deep Blue's 'expertise' possible. It just performs a vast number of calculations by brute force in a brief time.

-- Little Nipper (canis@minor.net), June 05, 2001.


Well I suspect that there's more to it than just "brute force calculations". I would guess that it uses some optimization techniques to narrow the list of possibilities. And I agree that given a totally unexpected situation, the AI would signal some alarms. Obviously there would be some manual override controls in the system. But AI supposedly "learns" from it's actions. I have no idea what NASA's software does.

-- Maria (anon@ymous.com), June 05, 2001.

I agree with LN Maria. Seems to me, that DB's chess ultimately reduces to memory, and non-biological computing power to predict all 100 trillion outcomes (or whatever) for a series of chess moves plus a man-made optimization algorithm and learning algorithm that enable robotic "thinking".

-- Lars (larsguy@yahoo.com), June 06, 2001.

Well, Lars here's some stuff going on in AI.

http://www.ai.mit.edu/research/abstracts/abstracts2000/newmodels.shtml

My comment was to point out that we've advanced it above "brute force", but of course it always will come down to ones and zeros. Can we advance far enough so robots can compose as well as Shakespeare? "Bicentennial man" and "short circuit" are just two examples of pure science fiction, which I think is well beyond machine capacity for ones and zeros.

-- Maria (anon@ymous.com), June 07, 2001.



To be or not to be, that is the trnhggflkuyo879erdazvczgfiuy&%@Y**#%#

-- (1_million_monkeys@the.steno_pool), June 07, 2001.

Moderation questions? read the FAQ