I have seen some radical ideas for attacking deep bunkers, but this beats ‘em all…
Having previously looked at Deep Digger and the Supercavitating penetrator, I was intrigued by an Air Force research Laboratory program called Creative Robots to Defeat Deeply Buried Underground Targets After finally getting clearance, I was able to interview Stephen Thaler of Imagination Engines Inc, the man behind the project. Thaler is evangelical about his brand of artificial intelligence, and the result is a piece in Wired News — “Experimental AI Powers Robot Army.“
Its quite a project. The idea is to develop software to make a collection of robots smart enough to break into, explore and neutralize deep bunkers. The challenges are gigantic.
The robots have to deal with an unspecified number of unknown obstacles as they travel via cable runs, air ducts, service pipes or other channels, dealing with grilles, bars, doors or other checks.
Then they need to correctly identify the target (waste bin, or WMD container?), which is easy for people but hard for robots and this task requires being out of radio contact.
They have to act in concert and help rather than hinder each other, co-ordinating their efforts to explore and map the facility.
And all the time they have to be able to avoid, outwit or defeat the human defenders of the bunker, whose tactics, numbers and abilities cannot be predicted.
Thalers believes his software can do all this. Its an unusual neural network with the ability to dream up new ideas, exploring likely approaches before putting them into action. For example, give it a set of robotic limbs and it will quickly find the most effective way of using them a video here shows a six-legged robot figuring out how to walk from scratch with no programming in eight minutes flat.
Imagination Engines capabilities also extend to sensors. Thaler describes products including a million-pixel array which can interpret input an order of magnitude faster than any comparable system and another with formidable powers of recognition, such as distinguishing a T-72 from an Abrams. There is no programming involved: just show the system the two different objects and it figures out how to tell them apart.
The most guarded aspect of the Creative Robots is their tactical intelligence, which seems to be considerable Thaler describes them as “Machiavellian” in how devious they can be. The Creativity Machine’s ability to explore the entire range of possibilities means that in principle it could dream up any tactic that a human could, and more besides.
Within the next few months the software toolkit for Creative Robots will be available for the military. It will run on any standard hardware, turning a pack of dumb robots into smart team players capable of carrying out missions on their own. Thaler believes their speed makes Creative Robots superior to those that rely on human control, performing at near-human levels of intelligence at Terahertz clock rates, while our joy-stick controlled robots are performing effectively at the 4 Hz clock rates characteristic of the brain.
The possibilities for civilian use are tremendous. There are a vast number of hard problems involved in getting robots to interact with the everyday world which require intelligence. Thaler believes that he has the solution. Look out for a host of commercial and industrial applications.
Dean Vieau, a consultant with many years of experience in the fields of Controls and Machine Vision, is an enthusiastic supporter. In one case study he carried out, Vieau found that a solution using Imagination Engines software was twenty times faster to develop and a hundred times cheaper than the existing approach.
Imagination Engines represents a significant advancement in the realms of AI. Not just esoteric academic conjecture but real world paths to concrete results.
As usual the military are developing world-changing technology that will filter down to the rest of us later. But are we really ready for killer robots yet?
There is a reluctance to entrust lethal missions to autonomous robots, says Thaler. However, the bad guys may not share the same reservations. The escalation is inevitable.
– David Hambling
Uhugg! I’ve got a bad feeling about this.
So what do these little buggers use for power? A real high pole in the tent of lots of neat concepts is batteries.
(In Steve Stirling’s last Draka novel, there was an unpleasant scene involving invasive robot bugs not unlike these. You don’t want them to get into your gut.)
I went to the company website, the bio of the founder listed research into consciousness and AI near-death experiences. Sounds more than a few steps into crackpot land.
“So what do these little buggers use for power? A real high pole in the tent of lots of neat concepts is batteries.“
This is a software project — the real ‘high pole’ in this case is getting something anything like intelligent enough to carry out the task.
However, having talked to AFRL people about their minitature robotics programs, they have a number of innovative solutions to the power problem. (Watch this space).
“the bio of the founder listed research into consciousness and AI near-death experiences. Sounds more than a few steps into crackpot land.“
Not if you read further. The ‘near death’ part is Dr Thaler’s rather colorful way of describing how his AI generates new ideas by peturbing an existing set of ideas.
As their website puts it ” Thaler found that if the connection weights were varied at just the right level, the network’s output units would predominantly activate into patterns representing new potential concepts generalized from the original training exemplars”…but saying that ‘a near-death experience causes the AI to hallucinate’ makes better copy for the media.
This is an embarrassing piece of “journalism.“
Did anyone notice that the experienced “consultant” (Dean Vieau) that he cites as having done an independent study on the company’s technology actually *works* for the company?
C’mon: put it all together, and it’s hard to deny the upshot of the deeply kooky aspects of Creative Intelligence’s website. Fine, “near death” may be creative marketing, but the hype that Thaler attaches to it means one of two things:
(1) he’s yet another AI or military scientist all too willing to hype his potentially mundane or fruitless experiments; or (2) he’s a kook.
Also, did you see the “In it’s image” links? Tell me THAT doesn’t REEK of kook.
Well, I’m not at all convinced that AI will ever work, but there is some potential here. I don’t think his robots will communicate or plan with each other any better than a 14 year old with his parents, but the identification software looks pretty neat. Distinguising an Abrams from a T-72 by sight, or learning to walk in 8 minutes looks interesting. I can see that a new programming technique could allow robots to do those things, and that’s nifty.
But AI robots that plan and create new strategies? Yeah, right. These guys aren’t Cyberdyne Systems.
Although, wouldn’t it be funny if we finally managed to create AI, only to discover that it was emotional and moody as a teenager?
Scientist: “We have created the world’s first artificially intelligent robot!“
Robot: “No one understands me! I hate you all!” (goes into own room, slams door, turns up stereo)
The most guarded aspect of the Creative Robots is their tactical intelligence, which seems to be considerable
No, it doesn’t scare the crap out of me. We haven’t seen any demonstrated ability here. For all we know, even if these things could think or reason, they’re just as likely to decide that attacking some terrorist is scary and they’d rather sit in an air duct, or watch TV, or that the best method to attack a terrorist is to jump into the trash can and deactivate themselves.
I find it strange that several posters just post ad hominem attacks on Stephen Thaler as opposed to actually taking the time to read what Dr. Thaler has published in peer reviewed journals. I have (for private enterprise) and the technology is a significant improvement in AI. The military is more open to trying unproven technologies than other large organizations so it’s not surprising the Air Force is conducting tests. The fact that they have distributed follow on grants, is further evidence that it works.
As the founder/operator of http://www.antimullah.com and also screenplay writer with a project called “History of the Future” I acknowledge the http://www.LindaSog.com post and link to this site a terrific find.
Whether my android/robotic fighters which develop over the next 3,000 years of future history or flesh eating locusts, as warrior or cleanup squads are all a figment of imagination they can be put to practical use.
Hmmm… any of you ever read the book “Prey” by Micheal Chrighton? Sound familiar?
How about the movie Screamers?
I know the TRUTH.
http://www.androiduniversity.com
Thomas Edison
The level of danger of a military robot software is limited to a timer. Ideal power source robot — to superflywheel 3D — suspension of its shell. Against the Chinese or Indian soldiers only robots. The moral of State will not allow the same loss.
http://www.androiduniversity.com Thomas Edison
I read:
I went to the company website, the bio of the founder listed research into consciousness and AI near-death experiences. Sounds more than a few steps into crackpot land.
Posted by: mike at September 16, 2006 03:48 PM
I have been interested in AI for 40 years so when I was in St. Charles Missouri I talked with Stephen Thaler at his offices located at 550 Wallstreet. He showed me some things he was currently working on and I was impressed.
I have been a software geek since the days when General Electric provided timesharing via teletype consoles.
My personal evaluation of Imagination Engines:
Great stuff!
Elmer Fittery
Principal Software Engineer