Can a machine think?
Thinking is a very general ambiguous term So, Turing asked the same question differently “Can a machine imitate a human?” which is widely known as “The imitation game”. Since then we had a test for machine intelligence in the form of the Turing test, which sees if a machine can fool a human interrogator into thinking they are talking to another human. However, the test is not without its problems and one of the famous criticism of Turing test says “The Turing test is like saying planes don’t fly unless they can fool birds into thinking they’re birds.”
Commonly, thinking is associated with intelligence – would this be a better fit? The dictionary definition of intelligence is the ability to acquire and apply knowledge and skills.
Computers are very good at acquiring information and making decisions based on that information, traditionally in a very fixed way. They can be programmed to do some specific task and just that task at the superhuman level, like playing chess. Deep blue might be able to beat any human grandmaster but it cannot do anything else even the simplest of task outside of its expertise. All it is doing is searching all the possible moves and choosing the best option available. Humans do not necessarily play chess in a similar manner. Humans rely on intuition and patterns whereas computer undergoes brute force search with some optimizations like Alpha Beta pruning. It is in a senseless confidence and has no knowledge of chess. It just knows the rules and it wants to win. It does not know anything about the chess. Yet, it can beat any human. Much like the way calculators can beat any human and airplane can beat any bird.
But this scenario has quite changed with the development of machine learning ideas. Recently, Google Deepmind’s Go-playing computer ‘Alpha Go’ beat the human champion Lee Sudol with the result 2-1 in computers favor. And earlier in the decade, IBMs jeopardy player ‘Watson’ shamed the human champions with a huge performance. Now, Go and Jeopardy is very different than chess. Certainly, the task is too complex to brute force. The computer can’t rely solely on their speed to search possible moves in order to perform any good. So, they had to be more like humans. They had to develop some kind of wisdom and intuition of the game. Therefore, they are modeled based on human brain with deep learning and neural networks. They were not programmed to play anything but were programmed to learn from an experience, with the hope that they will catch on some tricks of the trade. So, they were fed a huge amount of data from the online games database and other sources and they kept learning through experience in supervised or unsupervised manner. Now, this is interesting since they are not programmed to do or play anything but they are programmed to learn and evolve with experience. Alpha Go can play Go at superhuman level but it can also play other arcade games like Mario, Tetris or formula one at the superhuman level. These machines are not as dumb as they used to be. With enough experience, they can be evolved to do anything far better than a human.
Another demo of computer intelligence is the Google’s latest deep-dreaming research. Machines are creating abstract art – original pieces that no human has ever imagined. Some people might argue these images are more artistic than many humans could produce, but they are simply the result of a finite number of decisions and their knock-on effects as a result of training to see patterns.
The only drawback for the deep learning algorithms is the huge amount of data it has to be fed. One can argue that it is no different than any life. Species do not evolve in hundred years. It is the outcome of millions of years of evolution with each optimization getting stored in DNA and being passed to generations. So, evolution itself can be seen as the iterative process where the species get evolved and optimized with each generation much like the computer relying on deep learning with artificial neural networks. After all, the virus, amoeba, and bacterial life don’t seem that different from today’s algorithm. And each life up in the classification seems to be like the next layer of optimization, to the conscious humans. So, if our question is “Can a machine ever be conscious?” it doesn’t seem very unrealistic looking the way life evolved to this stage.
In this sense, humans themselves can be seen as the high functioning algorithms.
Bipin Oli
