Google’s AlphaGo defeated the worlds’ best Go player in a five round match of Go, a game known for its’ complexity and the seemingly infinite number of possible moves.
Another game conquered by IA – And not just any game, it’s a game where brute force computing power won’t cut it. Several techniques have been combined in AlphaGo, giving it a sort of ‘intuition’ that its’ developers don’t quite understand yet.
Although some would argue a computer beating a human at a game isn’t a development of any significance, many leading scientists see this as a validation of the progress made in the field of Artificial Intelligence in the last decade.
Aside from winning board games (chess) and America’s most popular game show (Jeopardy), AI has actually become much more prominent in our daily lives and will continue to do so at an increasing pace. Virtual assistants like Apple’s Siri and Google Now are using AI techniques to understand user intent and improve themselves by learning from user behavior and -feedback. Image recognition is used by Google to help you find your beach photos you didn’t label, or that friend whose face Google now recognizes.
In the near future a lot of these techniques will be combined to power self driving cars that recognize traffic situations and decide how to act in real time. And that’s just the beginning.
The ‘holy grail’ of Artificial Intelligence is general intelligence, which is something that transcends specific domains. AlphaGo for example, is very good at playing Go. But ask it to find you a good present for your friends birthday and you will end up with nothing. We might be a long way away from General AI, or we might be close. The thing with General AI is, once we’re there it might well accelerate to Super Intelligence. And that’s when things get scary.
Imagine a general intelligence with massive computing power and algorithms that can improve itself continuously and at an immense pace. The more it improves itself, the better it gets at improving itself.
Within very little time this Super Intelligent AI system will be smarter than humankind combined, and depending on the objectives and rules that were set in the system it can potentially do some serious harm to humankind.
A great quote by Eliezer Yudkowski on this topic:
The AI does not hate you, nor does it love you, but you are made out of atoms which it can use for something else.
Artificial Intelligence is a very complex subject, but also one that will most likely play a big role in our lives. A while ago Tim Urban (Wait but why) took the time to write two blog posts on this topic that explain how exactly this works. Tim has a way of making very complicated things seem simple, and he did the same with Artificial Intelligence. His blog is highly recommended, these two posts in particular:
If – after reading these two posts – you’re intrigued, then I would recommend reading Nick Bostroms’ book Super Intelligence. Nick has been researching and writing about Artificial Intelligence for years now and is widely regarded as one of the leading thinkers on this subject. His book explores the possible types-, likelihood- and implications of Super Intelligence. I never realized there was so much to it, so much to think about until I read his book.
