Slashdot Mirror
"Living robot" Escapes Lab, Makes It To...Parking Lot
jerkychew writes "This is either really cool or really scary, depending on how you look at it. According to this article, scientists in England have been experimenting with so-called 'living robots' that think and act for themselves. During an exercise that pitted the machines against each other in battle, one of the machines, named Gaak, was taken out of the competition and left alone for fifteen minutes. When the scientist returned to retrieve Gaak, he found that the machine had broken free from its 'cage', and made it all the way to the lab's parking lot before it was apprehended! Can the T-1000 be far behind?" Update: 06/20 20:36 GMT by T : Thanks to skywalker404, who points out the Magna site and Professor Noel Sharkey's web page.
Unfortunately we will only have the technology to enforce these laws many years after they have the potential to be broken.
All three laws are subjective and would require immense logic databases and analysis algorithims of constant environmental feedback imput. amazing how much the brain really does... not to mention the 'gut' whatever that is...
A fool throws a stone into a well and a thousand sages can not remove it.
Seriously, this is totally amazing. This thing:
- had the desire to break out of the cage
- did so and
- navigated to freedom
Needless to say, this warrants further examination. This sounds like roughly animal level intelligence. I hope they make more tests what this Gaak is capable of. It already sounds autonomous enough. Might this be the first step to true AI?
One thing to consider, though. Are combat and "survival of the fittest" type exercises REALLY what we want robots to base their intelligence on? It sounds to me like we are "breeding" them for aggression.
The small unit, called Gaak, was one of 12 taking part in a "survival of the fittest" test at the Magna science centre in Rotherham, South Yorkshire, which has been running since March
What better way to show your fitness than to sidestep the competition and make a break for it? Of course, poor Gaak didn't know about cars, or else it surely would have tried using the sidewalk on the way out of the compound...
What I'd really like to hear about is if this kind of behavior was duplicated in another instance. Otherwise I'd be very skeptical that this robot actually exhibited a "behavior" of wandering around trying to explore.
How much of this incident could be attributed to chance? Only by repeating it can you say for sure. I'm surprised this website is making such a big deal about this one incident when there's no proof/math/algorithms detailing an explanation of why this could've happened.
"In mathematics, it's not enough to read the words -- you have to hear the music"
I'm sorry, it just failed to strike me as anything major, simply because we don't know anything about the robots, the lab setup, the prior research or robot behavior, etc. etc.
All this means to me is that a robot drove out into the parking lot without anyone controlling it. Is that really so great a feat? I mean, if it is, please correct me here.
Do they know for sure that it was maneuvering itself towards the outside world with the actual intent of "escaping" or doing anything?
What would have been really interesting to see is what would have happened if they had just sort of followed it around outside for a day or two, of course making sure it didn't get destroyed or anything.
Palaces, barricades, threats, meet promises
From readingt eharticle, there seems to be very little evidence that intelligence was involved. What did breaking out of the paddock involve? Walking into the gate and having it swing open? How did it get out of the building? Did it find it's way, or could it just have "random-walked" its way out of the building and into the car park?
I remember playing with a toy that would change direction by 45 degrees or so whan it hit an object. It was also quite adept at finding it sway out of rooms, but there was no intelligence involved there.
----------------------------------- My Other Sig Is Hilarious -----------------------------------
I think there is nothing more to this than coincidenses and malfunction in the robot.
Will work for bandwidth
First Law:
A robot may not injure a human being, or, through inaction, allow a human being to come to harm.
How would we go about enforcing such a law?
In the Asimov stories, the First Law was somehow deeply ingrained in the mind of every robot's "positronic pathways" for the peace of mind of the human race. The fear was that the first robot to kill a human being would result in a mass destruction of the world's robots, due to what Asimov called the "Frankenstein complex".
But, welcome to the 21st century. In Japan alone, so far 11 workers have been killed by production line robots, resulting in precisely zero anti-robot pogroms.
We know, as technicians of the modern world, that the fastest, cheapest and easiest way to build something will almost always win. Our solution is not to write complex programs to give robot workers some sort of respect for human life, but to give the human workers around the robots a respect for the power and arbitrary nature of their mechanical colleages. Large yellow stripes are marked out within the working area of all robots, within which humans shall not go, and outside of which the robot (hopefully) cannot reach.
Of course, when you start giving robots wheels and independent goal-seeking behaviour, things get interesting.
Once more unto the breach, dear friends, once more, Or close the wall up with our American dead!
No he didn't - he made the three laws to show they WOULDN'T WORK, as he demonstrated in several stories.
For example, consider the first law. I don't exercise as much as I should. Since that will lead to ill health and death, a robot would be compelled to compell ME to exercise. No countermanding order would be accepted, since orders are Second Law.
Eat a cheeseburger? No, lots of "empty" calories and fat, little nutrition. That will cause harm - I must stop you.
Second law has its problems too, as Asimov pointed out. Bored punk kid runs around ordering robots to battle to their destruction for his amusement. Basically, every robot had to be given orders to ignore orders of self-destructive nature from anybody other than the owners, Universal Robots employees, and law enforcement.
Eventually, Asimov had to state that the three laws as stated were "fuzzy" - weighted by circumstances. Saving two convicted criminals is less important than saving one saint, obeying a foolish order less important than doing your job, etc.
Even that brought about problems - the incident when Hyperdrive was invented, for example.
Sorry, but should we ever create AIs, the most likely way we will be able to instill limits into their behavior will be the same as we do with people - years of training in "morality" and "ethics". Let us hope we get it right.
www.eFax.com are spammers
govt is already working on giving us equal rights to robots.
If it's talking cars it should be Michael, not Dave, surely?
erroneous: look me up in a dictionary
Actually the thing that scares me isn't the robot.
A bunch of scientists design and build "living robots that think and act for themselves", and all they can think of to do with them is to play BattleBots.
Man, If I were gAAk, I'd run away just to get away from those losers.
There is a group of monkies at the zoo. The monkies are kept in an environment consisting of a fruit tree, and there are a bunch of logs on the ground for the monkies to sit on, etc.
The monkies cannot reach the fruit, but one day, one of them props a log on end, climbs it, and just before it tips over jumps up and grabs a piece of fruit. The monkies had been there for 3 years and before that day no monkey had ever managed to pick fruit from the tree.
Now imagine taking that monkey and cloning him and starting the next day with all of the monkies having the insight to grab a piece of fruit from the tree.
Repeat this process day after day.
Slowly, you'd begin to select for intelligent behavior, and before long you'd have monkies that were far more intelligent than the starting group.
The point of the GA is to take the most successful members of a population and cross-breed them in order to discover the key elements of their success without positing a bunch of (likely incorrect) assumptions.
Amazing magic tricks
It was the fault of the victim, or some other human decision, that got someone killed or injured in every case you mention in Japan - and anywhere else in the world
It will always be 'in a way' the fault of some human decision if a robot does something. In a way, since we build the robots (or the robots that built the robot).
The reason there is no pogrom is that the robot was incapable of deciding to kill a human. The moment that becomes possible, and the first human is DELIBERATELY injured by a thinking robot, we WILL see an Asimovian response to intelligent robots.
This question relates to my last statement. When will a robot's action not be 'in a way' the result of a human decision? When will it be considered as a 'deliberate' action from the robot? Should it not be considered as the fault of the person who designed the robot (who designed the robot (who designed...))?
Can a serial killer defend himself by telling the world he was beaten by his father when he was young? Not entirely, but he will try. Why? Because, like robots, we are quite deterministic in our actions. It is always hard to decide who is taking the real decision; the creator or the creation.
phobos% cat
cat:
Ahhh, yes, like hominid group foraging. All these robots need is a decent sociology class, solar rechargable batteries, a big ass (auto-loader) weapon and we're in trouble. *snicker* ;-)
-- Thought is the essence of where you are *now*
Very cool story, but the media hype is, as always, excessive. According to the scientists in charge, it probably wasn't trying to escape but rather was "hallucinating" that random light (sunbeams) was prey, and tried to follow it out. Oh, and it stopped running away when the mixed sun/shade pattern of sunlight through a tree confused it.
Very cool, but I wouldn't consider it "frightening" in the way some media sources have been labeling it.
Mod me down, and I will become more powerful than you can possibly imagine!
(Somebody please mod up the parent post by MegaManX. I would if I could moderate AND post... anyone with moderator points remembers how hard it was to climb into the ranks of "bonus posting"...)
MegaManX wrote:
It will always be 'in a way' the fault of some human decision if a robot does something. In a
way, since we build the robots (or the robots that built the robot).
...
This question relates to my last statement. When will a robot's action not be 'in a way' the
result of a human decision? When will it be considered as a 'deliberate' action from the
robot? Should it not be considered as the fault of the person who designed the robot (who
designed the robot (who designed...))?
Can a serial killer defend himself by telling the world he was beaten by his father when he was
young? Not entirely, but he will try. Why? Because, like robots, we are quite deterministic
in our actions. It is always hard to decide who is taking the real decision; the creator or the
creation.
This is an excellent observation. I don't agree with it, but it does raise an excellent point, especially legally (a question for future robotics lawyers, I'm sure). Who is responsible?
The reason I don't agree is that we are discussing "intelligent" robotics. I suppose it would be generally accepted that a standard portion of the definition of robot intelligence would be the ability to make decisions completely unanticipated by the creator. If you program a system to learn from its environment, then modify its responses based on that learning, then the creator no longer has culpability for its actions.
(This brings up an interesting moral, or religious, point. The same could be discussed regarding God and mankind - does man truly have free will? Is he a product solely of his environment, or solely his Creator, or both? (I think both - I believe in free will.) It seems that every time robotic intelligence is discussed, religious overtones quickly arise.)
--Brandon / Split Infinity Music