One day, will the machines really become smarter than us?
Quite probably. If they become more intelligent than us, they will start to apply intelligence for their wellbeing. If necessary at the cost of others, including man. And they will come up with smart plans. In the present world, we do the same.
What is intelligence? Today's Wikipedia describes intelligence as the ability to perceive information, and to retain it as knowledge to be applied towards adaptive behaviors within an environment or context.
How intelligent are the machines today? And, how intelligent are we? Let's look at a few examples.
If you have opted in for Google Now on your mobile, it tells you the time it would take to go to work on weekday mornings (not on weekends) even before you ask. If you are somewhere on the road, near some restaurant or place of attraction, it pops up with details and reviews for it. It figures out your next travel plan from your emails and reminds you when to leave home for the airport. You can ask Google questions with your voice and you are very likely to get a precise answer. The fact that it can access the huge number of documents on the web may be attributed to sheer computing power, but the intelligence is in being able to answer questions spoken in human languages.
Let's talk about the autonomous cars that several companies including Tesla and Google are building. The cars can drive and park on their own under reasonably good road conditions. They can sense what's in front of them, can obey the signals and traffic signs and apply all the controls accordingly.
Dominos have started to use drones for delivering Pizza. Amazon is about to open a store where the customers only need to scan their mobile while entering and simply pick up the items they want to purchase and leave; their digital wallets or cards will get correctly billed automatically.
Man have built many other such intelligent machines. As of now, they are all built to assist us, to help us, as we think they should be. As of now, most machines or algorithms are designed for very focused tasks. That way, the machines can think only in the way we program them to think. They cannot connect the dots between two independently learnt scenarios by themselves. We can. So, we are far more intelligent than the machines.
But, why aren't the machines more intelligent than us? After all, our bodies are also biological machines following some mechanisms. Why couldn't scientists build machines similar to our brains?
Actually, they tried a lot, are still trying, and will keep trying. Neural networks, a computational model in machine learning, was invented several decades ago to emulate the human brain, with a network of neurons-like nodes and connections between them. However, the computational power of the machines could not scale upto anywhere near the number of neurons in a human brain. Consequently, neural networks were not successful.
The computers are more powerful now. Scientists are working with much deeper neural networks nowadays and building systems that can learn and apply much better than before. This new field of research is called deep learning. What can deep neural networks do? For example, they can detect properties of images much better. Try out https://quickdraw.withgoogle.com, and see how the system can detect what you may be trying to draw. Among many other interesting applications, deep neural networks can learn that the way Paris is related to France is the same as Delhi is related to India, just by reading a lot of documents.
However, even with the advent of deep learning, building machines smarter than human beings is far away. The number of neurons in a human brain is still way more than the scale of the deep neural networks presently. But, the way computing power of machines have increased in the last two decades, the machines may catch up by 2050.
So, it is not impossible either. The individual sensors are pretty much there, and are already very good. Cameras can already take better photos than we can see with our eyes, recorders can pick up sounds which we cannot. These sensors will improve further. Once the computing power improves by a few more orders of magnitude, simulating a deeper neural network as complex as our brain may become feasible.
We can argue that we do not understand our brain completely yet. But we will understand it better. Moreover, we don’t have to simulate the human brain to make machines smarter than us, it can be using some other model.
Still, it may not be possible to make machines smarter than us. No problem then. But, for now, let’s assume one day there will be machines that are smarter than humans. And if that happens, those machines may also have very complex brains. They may also have all kinds of emotions: empathy, kindness, ambitions, jealousy, ego. To some of them violence may feel good too.
The smart machines need not even look like us or the machines we are used to imagine and see in the movies, with a head, two hands and two legs. For example, an autonomous car can become a much smarter machine. It may become smarter than us, and someday it may tie us up in our seat and take us to some place according to it's wish, and not ours.
That’s what makes us afraid. We are used to being the smartest and the most powerful species on earth. If there are smarter machines, they can become more powerful. We are afraid that they will become monsters. We want to reserve the ability and right of being monsters only to ourselves.
But, we probably don't have a choice.
A. Whatever we do, smart machines will be made. As smart as possible.
Nothing happens by itself. People who are not doing it may be concerned, but people who are building artificial intelligent systems will try to make them smarter and smarter. Mostly, for our welfare. Eventually, if it is possible to make machines smarter than us, such machines will also be made. Ask yourself. Some day if you are in a position to make such a smart machine, will you make it or not?
B. Laws and regulations made by us will be of no use.
Laws and regulations are coming up to make sure innovation only sticks to the goal of human welfare. The only problem is: often it requires little alteration to convert something powerful and good into something powerful and evil.
For several years to come, we will practice research and development in artificial intelligence obeying some rules, regulations and ethics. However, those are just to prevent humans from developing systems which are not focused towards unethical practice or evil goals.
Such regulations will have any effect only till machines are in control of their inventors. Once machines become self-aware and can learn by themselves, they will be out of our control, just like most of us are not in control of each other. In fact, we will not even realize exactly when machines will go out of our control and will become almost equal to us. Maybe smarter than us.
If the machines become smarter than us, we cannot make laws for them. We eat chickens everyday. Can the chickens make laws for us?
However, it may not be all that scary. Not any more scary than it always was. Why?
C. We are anyway used to live in a world together with people smarter than us.
Would we mind if our children become smarter than us? Our students become smarter than us? We are used to and happy to produce people smarter than us.
The same is true for those who are not directly influenced by us. Are we the smartest in our society? Unlikely. Most of us know many others who are smarter than us. Most of us know of many famous or infamous people who are smarter than us. There are good people smarter than us, there are evil people smarter than us, according to whatever each of us perceive as good and evil. We are still living fine.
D. If smart people can develop smarter machines, those smarter machines can create even smarter machines.
Let’s imagine machines become smarter than us. Smarter than the smartest humans. That means, if a set of human beings could develop something smarter than themselves, those smarter things can develop something even smarter than themselves. That’s great. There will be a chain of increasingly smarter development then. There will be better doctors, better mathematicians, better artists, better philosophers, machines better in almost everything. I am sure we will like them.
There will also be smarter criminals, leaving aside the debate that defining who is criminal is often just a matter of which side you are. But again, criminals are not dumb. They have never been dumb. Society has always struck a balance.
There will be smart machines. We would just wish that they will be our friends, just like some human beings are our friends.
E. In the end, everything has an end.
Whether machines will be smarter than us is not clear now. Whether machines will destroy us is also not clear now. In fact, building smart machines are not the only way we can destroy ourselves or the environment as our habitat. There are several other concerns. Machines or not, we will anyway die.
In a few years, we will all die. Our children will die. Or maybe not, because people are trying to invent some technology to reverse our aging process. But as much as can be thought of now, the human race will die, even if we contribute to nothing to cause the end. The earth will be no more, the sun will be no more. With utmost certainty, we cannot be the only smart living species in this universe. Many other species must have lived and died in some other planets in some other solar systems. We are just very small living objects in a small planet of a small star of a small galaxy.
But then, each of our small happiness and sadness matters to us. Each moment we live matters to us. That’s why we live now, try to make better things, including better machines, and make things better.
That’s why today is called the present.
