We all read fantasy stories about self-thinking computers and human-like robots. Most of us are fascinated from the idea of a real Artificial Intelligent. However, what are the abilities that exist today? How far are we from this artificial intelligent? And what are the changes to actually reach it anytime soon? All the answers can be found in this article.
The idea of human-like thinking machine is fascinating humanity for several decades now. People think that it will be amazing to have a computer that thinks by itself and has its own personality. Most people, unaware of how computers really workare certain we are just a few years of research from reaching it. After all computers can do amazing things today, and most people do not understand how they do it.
However, the truth is that computers are not intelligence. They only reason why its look otherwise is because they are working according to a very detailed software written by intelligence people!
"Intelligence" computers today work in one of two ways: deep search or heuristic search, none of them is what most people will call intelligence or human-like. Both of those methods are based on the computer ability to make thousands of calculations in short time.
Deep search is the way of going over all possibilities that exist, calculating the final outcome from choosing this way and the possibility of ending in this location while choosing specific actions. Once the computer has all the values it chooses the actions with the highest probabilities to bring it to the best outcome.
Heuristic search is also searching the possibilities space, only here the people that wrote the software entered a few "experts' rules" into the algorithm. The rules detail situations or values that if the computer run into them, it will stop its search and will react according to the rules.
The best example is the Chess game. Nowadays computers can beat most (if not all) players easily. Most people think it means the computer is intelligence. Yet, the way the computer chooses its next move is very different from the way a human will. The computers will simply calculate all its possible next moves, the best response the player against it can do in return and so on, till it reach the end of the game and then it will calculate if this first move was a good one or not, based on what will probably happen in the end (the computer will win or loss). Humans cannot beat it sincehuman cannot calculate so many moves ahead and know the statistic outcome of each move. Humans need to use some other heuristic and intelligence observation.
Research made in the field
There is a lot of research in the field of AI, both in the academic world and the computer industry. This research tries to understand how humans work and integrate the good parts of it into the software (algorithm) that the computer uses. This idea helps making the computer more "intelligence" while still making it stronger than most human since it is rational and can use many complicated calculations.
However, there are many downfalls to this approach. First, since it just integrates human behavior into the software, it is a lot like the heuristic approach, which means that this approach does not bring us closer to the self-thinking and self-maintaining machine we vision when thinking about AI.
Second, this method demands to understand how humans will react. Therefore, it is usually researched in well define environments, and each such environment produce a completely different software (algorithm). I.e. unlike humans that can function in almost any given environment and adjust themselves accordingly, there is no multi-environment algorithm to solve problems. There are researches trying to make steps in this direction. However, they focus more on trying to build a universal environment for all problems, than building an adjustable algorithm.
Future of AI
How about the future? No one really knows. However, most experts agree that we are too far away from the Artificial Intelligence described in books, and we probably will not live to witness such a thing. The optimistic ones talk about 70 years from now as the first possible date. Most agree it will probably take longer.
However, there is one thing that might change everything and make it possible to have a real AI really soon. If there will be some kind of a new powerful development in the hardware that will enable us to simulate a human brain in the computer, we might have our "real" AI mush faster than predicted. Will that be a good or bad thing? – This is a question for a completely different article.