• Question: Is artificial intelligence as dangerous as the movies make it out to be?

    Asked by Holly to Fran, Peppe, Greg, Petros, Rumman, Pooja on 2 Nov 2017. This question was also asked by Nirvanaisthebest423, gary.
    • Photo: Giuseppe Cotugno

      Giuseppe Cotugno answered on 2 Nov 2017:

      Artificial Intelligence is a set of techniques which are used to resolve search problems (i.e. among all the routes, which is the fastest to go from A to B or among all the possible Go moves a computer can do, which one is the best for winning). Movies often show applications of AI to robotics where on top of “how few rooms I have to move in to find a beer” there are also non-AI problems such as “how can I move the arm without crashing into a furniture”. AI is a tool as a microcomputer can be. The microcomputers I work on can be used for operating a defibrillator or a cruise missile, the microcomputer in itself is not dangerous but people can use it to do dangerous things and the law should prevent this to happen.

      Movies are a fiction, as much as a book or a video game can be. They talk about a real fact but they add up something else to make the whole story spicy. There are a lot of movies on corrupted police officers getting busted by some colleague, but it does not mean that all police officers in real life behave like in a movie. AI is not different than that and you should be suspicious of all those business leaders, alleged academics and sorcerers that warn against the end of human race beaten by AI and machines. They are probably trying to attract some funding or sell some book. AI and robotics could be as revolutionary as it was the steam engine in the 1800s but it won’t autonomously decide to exterminate humans by itself unless another human would have told to do so (which probably has been attempted already when the steam engine was invented) and, no matter what, it will never violate the rules it has been told to follow

    • Photo: Petros Papapanagiotou

      Petros Papapanagiotou answered on 4 Nov 2017:

      Great question! At present, it is not.

      Many people have started believing it may become dangerous in the near future if we are not careful and take enough safety precautions. They believe computers will become smarter than humans (we call that “superhuman intelligence” or “superintelligence”) so they will start controlling us instead of us controlling them. Movies like Terminator, Robocop, and I Robot have plots like this.

      I personally don’t think this will happen any time soon. Although Artificial Intelligence has advanced a lot in the recent years and computers have become better than us in many smart tasks (playing Chess or Go), the programs that achieve this are very specialised. They can *only* play a particular game (or similar games to that). Building a computer that can be smart at *anything* like humans are is much much harder.

      However, we should be cautious about how we use technology. There is the saying that “computers make very fast very accurate mistakes”. We have reached a point where computers make so complicated and fast calculations that we are having trouble understanding why the result is what it is. For example, a program was designed to look at pictures of tanks and determine if they are American or Russian. It became really good at it, but when engineers analysed what it was doing, they realised that it wasn’t actually looking at the tank to decide. It was looking at the background, because most pictures of Russian tanks were taken at night, whereas most American tanks were photographed during the day. If it was a day picture the program would say the tank was American and if it was dark the program would say the tank is Russian. We thought the program was very smart, but it reality it was quite dumb. What if we use similar programs to help doctors and we think they are smart, but they really make dumb decisions that end up making people’s health worse?

    • Photo: Greg Chance

      Greg Chance answered on 5 Nov 2017:

      Current levels of artificial intelligence are limited to very specific tasks. Your calculator can probably do better then you in most complex calculations and Grand Master chess players have been beaten by AI inspired chess programs. Recently google programmed an AI to play old Atari games very successfully (almost perfectly) with a basic set of rules and rewards. These are great examples of where humans are beaten by machines. So the question to you is: do you feel threatened by these AI? There will always be some tasks that a device or computer or AI can do better than humans but this should not be feared, it should be embraced to allow us to engineer solutions to the real issues and problems of the world.

      So what about the future then? Will a highly developed AI be dangerous and want to enslave humanity? Well, first of all movies are there to excite, frighten and entertain you. Not to inform you of the development of technology. However, many notable figures, such as Stephen Hawking, have suggested we be cautious about developing very advanced AI. Alternatively, science fiction authors suggest the possible benefits to our society of advanced AI, giving us help when we need it. On this point my opinion would be that an advanced AI should have an advanced sense of responsibility taken from the best philosophical minds of our recorded history. If the AI is programmed to be benevolent and kind we should have nothing to worry about. And since we are programming them, we have no one to blame but ourselves!

    • Photo: Fran Zuch

      Fran Zuch answered on 6 Nov 2017:

      Currently not. Our brains are actually very complex and it is not easy to replicate it. So AI will not all of a sudden take over the world. As the other guys said, right now AI is only a very neat way for a program to make decisions and learn.
      At my company, we will mainly use it for client questions in the beginning. Clients do have a tendency to ask similar questions and for people who have to answer those questions day in and day out it can become very challenging to still sound enthusiastic 😉 Also, AI can answer questions at times when people in call centres are at home or on holiday.
      But that is it, the AI will only know and learn things based on the input we provide. It is still very limited.

    • Photo: Rumman Chowdhury

      Rumman Chowdhury answered on 7 Nov 2017:

      AI is dangerous if the people behind making it, using it and maintaining it are either intentionally dangerous or not aware of some side effect of it that would be dangerous.

Comments