 |
What is artificial intelligence? Artificial intelligence (AI) is the ability of a machine to learn and grow with time. It is an attempt to give machine the ability similar to human beings. Much of our machine are programmed or designed to do a specific work. But now stress is laid on making the machine smart and intelligent. But would our attempt to make machine smart create additional problems? We shall see in the succeeding paragraphs the weird problems that an artificial intelligence can create.
Firstly, they do what they are told to do. This might sound funny but artificial intelligence can do the things that they are told to do but in a weird way. For example if an AI humanoid is told to move from point A to point B at the fastest speed possible. It will move from point A to B that too at the fastest speed possible but it's movement would be kind of weird and not human like. So you can see how specific we have to go to make an AI do certain task.
Secondly, artificial intelligence can be programmed to do a specific task very well just like a machine. But when we want it to do all or most of the general tasks, they are unable to do so efficiently. And this beat the very purpose of artificial intelligence. We want machines to emote like us but we end up with machines which can't even do the normal work efficiently.
Thirdly, a machine is what it is fed and this could become a problem. As said earlier the machine is doing what it was asked to but what if the data on which decisions are made were not accurate. Just take the example of this one company which gave the AI the job to sort the resume of eligible candidates. It was doing fine until it started doing gender discrimination. It started rejecting female candidates’ applications. So was the AI biased against women in this case? No, this is not the case. It was later found that the company used to do gender discrimination at certain point in time and that data too was fed to the machine.
Fourthly, an AI is as good as the data it is fed. Suppose a car is programmed to avoid all sort of obstacles and accordingly it is fed with all the possible data. But suppose an obstacle come on its way that was never fed to it. Chances are, it will compare the new data with the existing data and think of the obstacle as an indication of clearance. This will make it treat the obstacle as a free path and thus lead to an accident. So, only because the AI was not fed with the data of one particular obstacle that led to an accident.
Fifthly, a glitch in the programme could make the AI behave in a way for which it was never designed. Maybe it will do the opposite of what it was asked to do. The AI can't be called rogue in this case. It’s just the errors in the programming codes that made it do what it did.
Sixthly, to make a true AI, it has to be fed with lots and lots of data. But what we want is the AI make sense of the data that it is fed, just like we humans do. But this ability would take time and efforts.
Seventhly, an AI will be rational and this could be a problem. We humans are emotional being and often make our decisions based on emotions. But an AI will think logically. Suppose you have to save your two kids trapped in fire. You would definitely save both of them. But an AI will calculate who has better chances of survival and will only save that kid. Teaching empathy to AI would be difficult.
Eighthly, suppose we make an AI that is almost human like. What is the guarantee that it will not revolt and demand for equal rights? Demanding for rights is still ok but what if it starts a revolution against humanity? What if they felt that they ought to be free and live free? What if some of them want to avenge their slavery problems, on humans, with violence? We can't really say. But yes the possibilities are endless.
Artificial intelligence is good but we must foresee the downfalls. It’s not that these problems don't have solutions, but we have to bear these points in mind before moving ahead with AI.
More