This is the most stupid thing I've ever heard.

"Robot may not harm...etc"

Robots aren't smart enough to know what is and isn't considered dangerous... For example, give ASIMO a knife and tell him to run forward towards a crowd. Is ASIMO going to hurt someone? Yes... On purpose? No... Is Honda going to spend millions of dollars and another 20 years trying to figure out how to avoid this? No...

These laws are just stupid. You have to understand that this is not as simple as it sounds. Even once you break the intelligence barrier to the point that robots can identify something that is dangerous, you will now have many new problems to face such as complexity. Once you have made a robot with human-like intelligence, you can't just *program* it to not hurt people. You can only *teach* it to not hurt people. For example, no persons brain is identical in the way we store things.

"Danger" for one person would be completely different to "Danger" for another person. It would be learned at a different time, stored in a different place, and stored differently, possibly have a different meaning and would re-occur hundreds/thousands of times throughout random areas in the brain (or AI in this case). You cant just plant a chip or program it to not be dangerous. Hence, what I'm saying is, once we have AI as complex as the human brain it will be impossible to take control. To deactivate such a specific thing in the brain would require a *complete* understanding of every single element stored in the brain, meaning you'd have to sort through possibly billions of unique symbols and decode every single one to see what they're meaning is, and once you've done that, you still have to figure out a way to program the robot to abide by such a thing, which would be a *lot* of work.

Hence, it will also be impossible to program a robot to not try to deactivate it's own deactivate switch. However, if the AI is done properly, the robots should be no more harmful than the average human being, so there shouldn't need to be a deactivate switch or any worry for that matter.

If the robot is proven to be unsafe, then the robot shouldnt be allowed in public places without the owner supervising it (or not allowed in public places at all depending on how unsafe it is). Installing an emergency shut off switch will only cause more problems and be targeted by theives and pranksters.

Only the owner of a robot and high authority should have the ability to deactivate a robot, not the public. My recommendation would be a remote control system that requires a pin code to activate/deactivate the robot. If the robot goes crazy, the police can identify the robot, obtain the pin from the owner and deactivate it themselves.

Japanese government should spend less time watching movies and getting ideas from science fiction novels and open their eyes to the real world of robotics and AI.

Source: I study brains and program artificial intelligence for computer games and robots.

Just a few more things. Look at the human brain for example. Our brain is very well made, and it is very intelligent. Our unconscious mind can force things upon us. It can force us to not do something such as commit suicide or murder, and yet, even individuals who don't want to do these things (and are stopped) are capable of doing so. There are also many ethical issues that can change certain circumstances, for example: What if killing an individual was necessary in order to protect the rest of the population?

I mentioned earlier about the complexity of AI and the possible billions of symbols. It's not just decoding these symbols that becomes a problem, it is also the fact that all these symbols are somehow connected to other symbols. It's like the internet in a sense. Many things are connected, and some connections are very important, whereas others are not. If you start manipulating with these symbols and their connections you could possibly make things much worse and create serious problems that may not be visible until a specific event occurs, and you won't be able to desk-check (test) if you've solved the problem completely as you would in a more basic robot, because it would simply be too complex to do so, so you would have to rely on a different form of testing (physical testing) which would disguise many bugs in the altered AI.

On another note, if we leave the subject of standard robots and move on to androids, these laws could (and most likely will) upset the androids. As soon as androids are given emotions, feelings, beliefs etc. equivalent to that of a human, the androids (like ourselves) are not going to be too happy about the fact that they are under our control. Put yourself in their shoes for a moment. How would you feel if you had a huge red 'emergency shut off' button on your forehead, and were under control and surveillance by another (seemingly-equivalent) species? You wouldn't be too happy. The best solution therefore would be to treat androids and humans equally in society (this is assuming that the androids have reached the same level as humans in terms of intelligence, emotion etc.), otherwise we will see a divide in species (human and android) which could have serious consequences.

While androids with emotions may seem like a joke to some, it's not. It's very possible to do, has many benefits, and will definitely be in cooperated into the future of robots/androids. Emotion (like intelligence) is all about algorithms. So if the human brain can do it, a computer can do it too. So don't say you weren't warned.

Here's a highly debatable question. If an android goes haywire and kills a human being, who is to blame, the:
* robot?
* owner?
* store?
* manufacturer?
* government?

If we assume that the manufacture did not create a robot with any faults, and that the government did not create any faults to the AI through stupid laws, then the question would be equivalent to, "if a human shoots another human, who is to blame", the:
* human? [robot]
* the person who provided training on how to use a gun? [owner]
* the person who motivated the murderer to murder [owner]
* the person who sold the gun? [store]
* the gun manufacturer? [manufacturer]
* the government which legalized the use of guns? [government]

It may be a surprise, but the manufacturer would not be held responsible (and shouldn't be). It would most likely be the owner. However, is justice to put the blame on the owner when in fact the android has the same intelligence and emotion of a human being? Who is to say the android did not go against the will of the owner?

Anyway, sorry to scare anyone. It's really nothing to worry about. Todays robots are very stupid and rely on doing specific tasks that they were programmed to do (unlike humans which are made to learn, understand, solve and make decisions etc.). No one has made a truly intelligent robot, or come close, and it is impossible to predict how long it will be before we see such a robot. In fact, the problem of understanding how the human brain works is so hard that we have never been able to create a theory on how it's done (something that could have been done even before computers existed).

It's useful however to keep in mind, that even in the future, androids shouldn't be considered as 'evil' nor 'nice'. They will be equally as unpredictable as any ordinary human being, so it's only wise to treat them as exactly that.

Manufacturers on the other hand, should look at this new law as a guide or message stating the importance of safety, but it shouldn't be taken it too literally as not everything you see in movies or read in science fiction novels is practical or possible, and in this case, I would assume the government only wants what's best for the public.

The only way to really make a safe robot (today) is to test it under many scenarios and programming it to identify certain things that could be considered as unsafe, and, of course, the design and capability of the robot would also have to be taken into account.

In the future, robots will be taught what is safe and what is unsafe, and the robots (which unlike todays robots) will *understand* why something is safe and unsafe and will be able to figure out things on their own. So technically speaking, the next generation robots (e.g. the truly intelligent kind, equivalent to that of a human) will actually be *safer* than todays un-intelligent robots.