Black Mirror and Turing Test

I have been watching the pilot of different Netflix’s series lately and last week I watched, by mistake, the first episode of the second season of Black Mirror. If you didn’t watch it yet and don’t want any spoilers, stop reading this right now and come back once you finished it.

For those who stay: I did enjoy it. It is called Be Right Back and it is about a couple in the near future. It is an ordinary couple, overall happy as most. That is until the man goes out to buy something and dies. During the funeral a friend of the woman suggests a neat service: a chat that mimics the deceased ones by answering questions based on everything they posted on the Internet.

It is the kind of science fiction which sends chills down your spine. That is because it is actually very realistic. There is nothing on it that can’t actually happen in the not so far future. Nothing actually impossible, although some things, like an android you grow on your bathtub, might be highly improbable.

I remember reading an article a couple of years ago about a software that would recreate 3D buildings doing exactly that – collating pictures posted on the Internet. If anyone remembers it or has the link to it, feel free to post here.

However, if that was already possible to do, one would have a winner of the famous Turing Test here. One might say that looking in the Internet for realistic answers would be cheating, but I am not sure I agree. In a sense, what one would be proving is that the combination of an algorithm plus the Internet passed the Turing Test. But the internet is nothing but a repository of collective memories. Saying that using the Internet as a source is cheating is the same as saying that with a large enough database of answers and questions one can pass the Turing Test. In my experience with machine learning, the size of the database, in fact, poses a different problem.

The larger the database, the more difficult is to memorise. In other words, the program would have to be infinitely fast to search a very large database and answer the person at the other end fast enough to look like a human. Because there are limits to how fast search engines can be, it would have to make choices, which is part of being “human”. In addition, it would have to not fall into traps like repeating the same answer to the same question and, actually, notice if its partner is repeating it too much.

Still, if you adhere to the functional definition of consciousness and believe that something that pass the Turing Test deserves its rights, it would not be very improbable that sometime in the future we might be forced to discuss whether the Internet is alive or not.

That was good science fiction!

Leave a Reply

Your email address will not be published. Required fields are marked *