Climbing The Mountain: The long foggy race to General-Purpose Superintelligence
If you want a metaphor for AI as it is now and has been for a while, imagine this: All the big players in the tech scene are running against each other in a giant race. Everyone is running at a leisurely pace, passing each other every so often. Its a hard upwards climb and the path is foggy.
The end is unknown; it could be another 10 miles, 20, 50... no one knows, in fact, there isn't even any guarantee that there even is an exit. they are just running the best they can, trying not to get completely worn out.
As soon as someone announces "hey I see the exit", they will all make a mad sprint for it, because the stakes are high, they don't want to lose when the endgame suddenly becomes very real.
So far AI has proven itself very useful in a wide variety of tasks, but it has always been restricted to the specific tasks it has been programmed for (with very few exceptions). The holy grail of AI, and the original grand vision, remains illusive; A machine that can learn totally new tasks it has never encountered before, without being reprogrammed, like the human brain can. A general purpose AI or AGI - Artificial General Intelligence.
The coming investment explosion (the sprint to the end) before the intelligence explosion
I predict that when someone does demonstrate an AI with very basic general purpose intelligence, the tech world will be thrown into a positive feedback investment cycle that will launch the general AI technology to superhuman levels within a relatively short space of time.
Think about it; would you want to be Microsoft knowing that Google could potentially be about to develop a super AGI and you are not? You would do everything in your power to make sure you are well ahead of where you think other others might be.
We could see super-human general intelligence emerge in an little as a decade or so after someone makes a basic general intelligence, due to the explosion of funding the field will receive.
My money is on Google, they seem to take the Singularity stuff more seriously than any other giant company. Its possible that general intelligence could come from a startup, but existing big companies will be too keen to acquire promising machine intelligence start ups as soon as they have something new and really impressive to show. The obvious example of this is DeepMind.
Deep learning is a powerful first step on the road to generating basic general intelligence, the same basic algorithms can teach itself to play all kinds of basic computer games, to understand language, (read and write), and to understand images and video.
In fact Deep Learning may be The only example ever of something hinting at the possibility of basic machine general intelligence. It is the first spark of life in the coming revolution. It still has a long way to go before demonstrating something that could be comfortably called an AGI but things are moving quickly.
The next few decades will be key. The AIs are coming.