-
Contents
- Sweet Confusion: Does high fructose corn syrup deserve such a bad rap?
- Closed Thinking: Without scientific competition and open debate, much psychology research goes nowhere
- subscribe
- give a gift
- renew
- RSS Feeds
- email alerts
- Digital Edition
- Podcast
- Past issues
-
Contents
- Better chow yields more milk: A more nutritious form of corn for dairy cows boosts farm profits, teen investigator finds
- Fuzzy future: Kids may suffer impaired vision from spending too little time outdoors, studies suggest
- Building with moon rocks: Working on the moon with lunar soil and grit could prove easier, more efficient and less costly than using earthly materials
Web edition: September 24, 2010
Print edition: October 9, 2010; Vol.178 #8 (p. 32)
As director of the Maryland Robotics Center, Satyandra Gupta oversees 25 faculty members working on all things robotic: snake-inspired robots, robotic swarms, minirobots for medicine and robots for exploring extreme environments on land, under the sea and in outer space. In September the Center hosted its first Robotics Day; afterward, Gupta talked robots with Science News writer Rachel Ehrenberg.
How do robots influence our lives today?
There are certain scenarios, such as manufacturing — making cars, making airplanes — where people are replacing human labor with robotic devices and the rationale is usually that it is less expensive, quality is consistent, that kind of thing. Then there are certain applications where very few humans can do the task because the skills required are so high…. Surgery would be an example. Let’s imagine that there’s a very hard-to-perform surgery that very few humans can do. Now if a robot can be trained or even teleoperated by these surgeons, then you would be able to get that performance from that robot.
A third scenario is where humans can do it but it is so dangerous that it doesn’t make sense…. Disaster relief, mining — if people get trapped and you want to send somebody in, the situation is often dangerous and rescue efforts can get delayed quite a bit. Another category is in space applications. Sending people to Mars to do explorations, it’s just not feasible. You don’t have any alternative except to send a robotic vehicle there. Or if you are going to send something to a planet far, far away and it takes, say 10 years to get there and 10 years to get back, it’s just not realistic to think about deploying humans.
Also, for example, with prosthetics, these are tasks where no human can be the replacement. If somebody lost a limb and you need to give them an artificial limb, then there’s no competition, no human substitute for that, for the capability needed by that person. Similarly when you are looking at a very small airplane doing surveillance, you don’t have an option of a human doing that surveillance because, in order for a human to be sitting in there and flying it, the size has to be bigger than a certain threshold. But if you have a very small plane which can do the surveillance for you, then you have all kinds of advantages and maneuverability. So there the robot is not a substitute for the human; the robot gives you brand new capabilities.
Is it important to have thinking robots?
I think the goal has been to make the robot intelligent.… The idea being that you should be able to give the robot a goal, a mission, and the robot should be able to take actions based on the mission and whatever information becomes available. So that’s what people would like, as opposed to programming a robot at a very, very low level that specifies move this way, move that way.
Isn’t it hard to mimic what was created by billions of years of evolution?
Yes, at both levels. Mimicking evolution in the physical manifestation is hard. What tends to happen is natural systems have lots of little joints, lots of little muscles which pull and tug and make the motion. And if you start introducing that many joints and that many motors in a synthetic thing, your cost goes up exponentially as complexity keeps growing. At the same time, the reliability goes down. If any single joint fails, your whole thing is going to fail. So that makes it very hard to mimic nature.
On the brain side of things, it is also very, very hard. A lot of us teach ourselves how to reason, how to think, how to analyze new information and make sense of it. This has been very difficult for robots to be able to do. So people will try to program different contexts and different scenarios for what a robot should do, but that way of doing it is simply not scalable. You don’t have enough time and manpower to code up all the possible scenarios that a robot would encounter and what it should be doing. People have made very small steps toward reasoning and learning, but in general robots do not yet know how to learn.
Will we ever have to worry about evil robots?
I’m not too worried about that, but I am really worried about the virus part of it. All of us have computers, and there are some people that are intent on infecting our computers with viruses. So in that way, you will have a bad or evil robot. Not that it thought and then became bad or evil, but it will be evil because somebody wrote a virus and infected its brain.
Please alert Science News to any inappropriate posts by clicking the REPORT SPAM link within the post. Comments will be reviewed before posting.
I reject the idea that computers can solve all problems. Turing showed they can’t, but computing is held in religious esteem by many, worldwide. I grant that computers are very good at processing data according to rote procedures, or algorithms, which was Turing’s original intent. When used to monitor and control physical processes, however, computers, microprocessors and microcontrollers leave a great deal to be desired. It is a shame that strong AI and other lofty projects must strive to work through such a restrictive mechanism as the Turing machine.
It is not a question of whether the Turing paradigm “works,” given many existing examples, but how difficult it is to practice that technology and if the outcome is always appropriate. The computational model isn’t feasible without software, but the fundamental problems of software persist even after six decades of experience. The necessity for software complicates the hardware requirements, makes the management of temporal concepts difficult, and prevents true parallel-concurrent operation, among other impediments. For my list, see Impediments of Computation on scribd.
Computational language and process is limited to transformations and translations in the space domain. We contort ordinary problems in order to make them suitable for computation. A common means is to convert all parameters of a control problem into sterile data by sampling, after which the only method that can work is computation (mechanical manipulation of data). When your only tool is computation, every problem, task, or job begs for data translation or transformation. If not intrinsically so, we tend to convert tasks into computational problems because the computer is ever-present and ready-to-use. The pre-task becomes, “How can I arrange this information so as to take advantage of the available computational resources?”
Other methods can be devised, but it takes some thought.
I have approached generic control problems in a very different manner and have invented and developed Natural Machine Logic, a non-computational method of automating processes using stimulus-response temporal hardware logic elements. My new language of natural logic has eleven root-words (operators) which encompass the native operations in space (Boolean) time (Moeller), and (joint) space-time (Moeller).
Living temporal awareness does not require a clock for its perception. For a sense of time, one need only to notice that events are related, one to the next. The “befores” and “afters” indicate the workings of nature and of man. Causes occur before effects. Consequences occur after actions (or failures to act). Most telling, it is the order of the notes that determine the melody; the order of the steps that make the proof. It is order, more than duration or strength, that determines the process. It is the understanding of processes, and the ability to devise them, that enables man to rise above mere survival. Order is the key to process.
If order is so important, why do we not have a logic for it, a proper temporal logic that is not couched in space? The formal logic given us, descended from the masters, admits only existence (both presence and absence aspects) and coincidence (in both space and time).
My system of logic is much more expressive and will eliminate, in time, the necessity for most control-system run-time software. Work with NML has generated two patents, one magazine article, and several conference papers on this unconventional theme in past years, but the major portion of work on NML has not yet been published. It is so tough to buck the interests vested in software and conventional thought.
I need help to take this new technology to market and am willing to share its benefits.
Any advice will be appreciated.
c.moeller at ieee.org
That was a really great summary of the status of robotics! I learned more in your one page than I have from many long articles on the subject, mostly because you seem to have a knack for bringing out unexpected implications in a clear way. I look forward to seeing more such articles by you!
Cheers,
Terry Bollinger
You must register with Science News to add a comment. To log-in click here. To register as a new user, follow this link.