When they say the future is AI, one of the most common reactions among people is that kind of is impending hopelessness, that is partly based on cynicism and partly on fear that we humans will become obsolete. While the reaction is somewhat understandable, let’s get one thing out of the way – the end goal of AI is not to replace humans and mark them redundant, but to make life easier for all of us so that we can devote our time and energy in doing things we want rather than doing what we have to.
And that is why one of the most exciting turns AI is taking is to work towards a state where machine learning walks hand in hand with neuroscience. Our brain’s design might be the most efficient way for the creation of local intelligence. Local, because the direction deep learning is taking, especially from what we can see through projects undertaken by Google DeepMind, is that the next step in the evolution of these things is building AI that is not just built for specific tasks but ones that serve a general purpose. These are designed to learn new and complicated things not through crunching TBs of data but sort of the same way the human neural network observes, processes and learns.
You could say that Apple’s Siri, Google Now, Microsoft’s Cortana and Amazon’s Alexa – they started out by accessing huge troves of data and running thousands of permutations and combinations in split seconds, but the AI race is now heralded by Google’s AlphaGo, thanks to DeepMind, which is now able to recognize a thing by seeing it only once, similar to how the human brain works. And this is all made possible due to constant advancements in deep learning, where the mind and the machine are always converging to create a ‘solution for intelligence’.
The reason the phrase ‘general purpose’ is so important in deep learning is because, the idea is to build something that can take a lesson from activity A and then use it in tasks B, C, D. The deep learning it gets from say mastering the ‘80s arcade game Space Invaders, it’ll try to solve the more complex problems of a more advanced shooter, say the latest Call of Duty, that too against the best human players. AlphaGo has already beaten the world champion of Go, the traditional Chinese board game. That is more pivotal in the history of computing because the number of possible moves in a typical game of Go is millions of time more than a game of Chess, which itself is very complicated. In such situations, we need machines that can actually use human-like intuition instead of a super fast number cruncher.
Bottom line is, the merging of machine and mind is only the natural step of the evolution of tech, and while it’s dependent on machines, the innovations and breakthroughs are spearheaded by real people. That’s something to think about.