Alan Turing, an English mathematician who lived in the middle of the 20th century, was one of the most influential figures in the development of computer science, in the formalization of the concept of algorithm, in the creation of the modern computer and a pioneer in artificial intelligence. He is known as the father of computing and a little bit of his story was told in the movie “Imitation Game”, in which Turing is played by the celebrated actor Benedict Cumberbatch.

One of its main legacies is the famous Turing Test, used to prove the artificial intelligence of a computer.

The illustration above is a representation of the Turing Test: the man in the middle is alternately conducting a conversation with a living human being (on the right) and with a computer running an artificial intelligence program (on the left). Alan Turing said that if the interrogator fails to identify which of the two participants is the human or the machine, the computer is declared to have passed the test. The test result does not depend only on the computer's ability to give correct answers to questions, but mainly how similar the answers look to an answer given by a human.

Virtual assistants commonly referred to as “chatbots” are making great strides and it is very likely that you have interacted with one of them on an e-commerce site or your bank, without even knowing it.

The technology behind these virtual assistants is known as Natural Language Processing which basically consists of 1) computer understanding of written texts through the analysis of phrases and words and their syntactic meanings, semantics and morphology and 2) generation of phrases and words by the computer that have meaning and similarity to the idea that is intended to be represented.

The truth is that although some virtual assistants are quite convincing (for example Google Duplex), so far none of them seems to have passed the Turing Test. Alan Turing himself predicted that in the middle of 2000 computers could pass the test, but according to several experts the expectation is that it will take another 10 years, that is, in the middle of 2030.

So far, these virtual assistants have achieved a good result for short and direct interactions, for example, to answer questions such as: "What is the status of my purchase order?", "Would you like to book a hotel room", where ambiguity is small.

When the interaction becomes longer, more like a normal dialogue between people, the results are still limited. The reason is that human communication is very complex and is not limited only to the content that is transmitted by verbal or written means, but many other nuances such as body language, tone of voice and humor, the context of the conversation and the personality of the interlocutors. Also present are regional and cultural factors that influence jargon, slang, accents and the degree of assertiveness of the conversation. In addition, the triggering of subjects in human conversation rarely follows a linear pattern, related themes or cannot come and go suddenly.

So in order to really allow a longer conversation or a more natural and fluid dialogue, virtual assistants still need to incorporate intelligences that map the various contexts of the conversation and be prepared to access different areas of knowledge depending on the themes that may arise. In the latter, the Knowledge Graphs stand out, which are data structures that allow the aggregation of knowledge, coming from different sources, which have relationships with each other.

Since I started with Cumberbatch, I will conclude with Harrison Ford: in the first Blade Runner (the second is also great), the test applied at the beginning of the film, called Voight-Kampff, is a fictional derivation of the Turing test and aims to detect replicants - human-looking robots designed to replace man in dangerous tasks. Want to know if you are one of them? See this link: Are you a Replicant?

Are they among us?