In this video Viktor Dörfler asks whether machines can outsmart humans.
In this video I tell a few stories of machines outperforming humans in areas that are supposedly uniquely human such as chess and go. I also show how this also happened the other way around i.e. humans defeating machines, and when this has not happened, that I expect that it will happen soon.
But I also say that I believe that in these areas machines will eventually be victorious. However, the reason is in the nature of these areas: they are well-structured (at least, they appear to be). In such areas data processing can potentially outperform thinking. I am not afraid of the smart machines outperforming humans, I am afraid of those who want to reduce thinking to data processing. Thinking is so much more than that. I heard a Google developer claiming that they had constructed a programme that understands the concept of the dog. He explained that he made the machine learn so many pictures of dogs that if you showed it a picture of a dog or one of a cat, it could recognise the previously unseen dog as a dog, while would say that the cat is a non-dog. I’ve asked if it can be happy (or angry, does not really matter) if a dog licks its face, if it can pet a dog, or get scared of a really aggressive dog. If not, it does not understand the concept of dog. The developer did not understand what I was getting at.
This reminded me of a story about John Lennon
: Supposedly, when he was five, he was asked what he would like to be when he grew up. He said happy. They said he did not understand the question. He said they did not understand life.
Credit: Copper engaving of ‘The Turk’ by Karl Gottlieb von Windisch’s 1783 (Public domain via Wikimedia Commons)