Here is quick history of the AI field. We need to creating thinking machines, but we dont know what is thinking so lets take any human activity that requires t

Here is quick history of the AI field. We need to creating thinking machines, but we dont know what is thinking so lets take any human activity that requires thinking, oh chess is such a game, lets build algo to play chess, yeh it works and beats humans, but can we say it thinks? hmmm hmmm not really coz this algo can't play any other game .. dammit this is not AI. We are stuck. We need to design algos that can do a lot of different tasks and not just one like playing chess also these algo should be able to extend their capability rather than redesigning them. Hmmm lets use general purpose "search" algos (LOGIC programming) and also feed lot of "knowledge" into the algo as input. Yay we got expert system and they almost looks like thinking. Let's see what else can we use them to do and someone shouts "lets see if they can recognize image", dammit that's so hard to "represent" using "knowledge bases". We are stuck. Lot's of people kept working on how to "represent" the algo and inputs for the problem of image classification and similar problems. Came up with various "machine learning algos" including "deep learning" algos which seems to work really really well but of course they have their own demands like huge amount of data and computing power etc. But are they really "thinking" ??? They can be easily fooled, they are like black boxes nobody can extract some useful knowledge from trained network such that it can be used somewhere else just like general purpose knowledge we humans have. This is the latest shiny thing as of now. But we don't have general purpose AI yet. We always want silver bullet and we always hope that we will get there by creating hype but in the end the hype dies and people move on to next hype.