First Published as part of the now discontinued blog TMTalks.com 25/10/2017.

Who am AI?

True Artificial intelligence is almost here! Or is it? I don’t know. I am not sure we will ever know. It is too difficult to pin down exactly what is meant by ‘intelligence’. For us to try to understand this, let me take you back a hundred years, to a time before ‘thinking’ machines. Now, imagine if I were to tell you, “one day in the future we will create a synthetic machine capable of beating the greatest human minds at chess”, you would probably not have believed me and said that such a machine must be extremely ‘intelligent’. Fast forward to today and we have machines capable of doing just that. Do we think they are ‘intelligent’? No, of course not, they are just programmed very well and follow a strict set of rules. Which leads to the question: what is artificial intelligence?

Developed by Alan Turing in 1950, the Turing test was designed to assess a machine's ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human.  The crux of the test is to see if a human interrogator can reliably distinguish between communicating with a man or machine. The test attempts to prove if a machine mind is on par with a human’s.

However, the Chinese Room experiment throws doubt into the equation; it asks you to imagine yourself as a monolingual person in a locked room with nothing but a rule book, some paper, and a slit in the wall. The slit in the wall is used to pass you a sheet, upon which is a string of Chinese characters. You recognise the characters shape, but do not know the meaning of the text. By using your rulebook, you write the corresponding string of characters defined and pass the sheet back through the slit. The process is repeated, and a detailed interaction takes place. From outside the room it is impossible to tell how the replies are generated and that we are not just communicating with a Chinese speaker. However, although the answers you wrote were flawless, you never understood the meaning. Following this chain of logic, no matter how much a computer imitates intelligent behaviour and no matter what programming makes it behave that way, since it cannot understand the symbols it processes, it's not truly intelligent, it’s not actually thinking.

But what if we could create an artificial intelligence that had the ability to understand? One that could ask who am I? It opens a whole new realm of possibilities and difficult questions. Can an AI be evil? you wouldn’t call a kettle or a watch evil, surely a ‘device’ is evil by design. But what happens if an ‘evil’ AI is designed by a previous ‘good’ AI? Who is responsible?

What do we do in this world, if not ask difficult questions. Mankind has an uncanny knack for asking questions and then pursuing the answers. It is probably this constant state of searching for answers that we consider to be a defining feature of our intelligence.  But rarely have the answers we have pursued held such magnitude. For me, artificial intelligence is a machine’s independent intelligence, its ability to conduct a complex task with limited input from the user. I struggle to see computers, built on pure logic and reason, developing original thought to the point of philosophy and an understanding of self-awareness. Having said that, the neural network robots behind google translate have already created their own language which is 'not readable or usable by humans’. Perhaps soon, it will be us not understanding the machines, rather than the machines not understanding us.