The turing test
Who are you chatting with - a friend or a machine? See if you can recognise who is which in this stand.
What do we see in the stand?
In the stand you get to chat with someone - but is it a human or a machine? It is difficult to explain what intelligence is, but easy to recognise. This was the point of departure for mathematician and pioneering computer scientist Alan Turing when, in 1950, he proposed a method for determining whether or not a machine is “intelligent” as we understand the word. A human interrogates the subject – who may be a human or a machine – using a text-only channel. If the interrogator is unable to determine that the subject is in fact a machine, one can say that the machine is intelligent. Here, you can chat with someone who may be a human or a machine (the choice is random). Can you determine whether you are conversing with a person or a machine?
What is Uppsala University doing in this field?
Uppsala University has been conducting research into artificial intelligence (AI) for many years. This includes research into machine learning – computer programmes that can learn by example in order to, for example, interpret images or natural language. Among other things, researchers at the Department of Information Technology study how computers might mimic human intelligence, including drawing conclusions from images and other data or learning tasks from experience. The Turing Test is a classic means of determining machine intelligence; if a person involved in a discussion is unable to tell whether the other party is human or machine, one can consider the machine to possess intelligence.
What are the benefits for society?
There are many situations in which it would be extremely useful to have a computer that can perform tasks that are difficult to explain precisely and that would otherwise require human experience. This might be something as mundane as the camera in your telephone automatically processing a photograph you have taken to give the best possible result, or it might be in more technical areas of use, like self-driving vehicles or autonomous diagnosis of cancer from X-rays. As the solutions to these tasks are difficult or impossible to describe in exact detail, artificial intelligence and machine learning are ideally suited.