Hey my friends, this is the article 76 / 1000! For those who don’t know me, my name is Selim Chehimi and I’m an engineering student. I’ve been programming for 7 years now so you can tell that I really enjoy that. Those 7 years of programming lead me to Artificial Intelligence. My dream is to build an AI Startup so that’s the reason why I’m sharing this article with you. I hope you’ll enjoy reading it as much as I’ve enjoyed writing it.
One of the major events in AI was when Deep Blue beats the world chess champion Gary Kasparov in 1997.
Gary Kasparov vs. Deep Blue
However, Deep Blue is what we call a Narrow AI which can only do one specific task (in this case playing chess). It can’t do anything else. From an AI point of view, this is unsatisfying because it’s not that intelligent. Deep Mind wanted to go beyond that Narrow AI so they came up with a new technique called Deep Reinforcement Learning.
Deep Reinforcement Learning
Deep Mind is working on this new technique since the beginning of the company. Deep Reinforcement Learning combines two AI techniques: Deep Learning and Reinforcement Learning.
Deep Learning is used to allow the AI to perceive the world around them through neural networks. Reinforcement Learning consists of selecting the right action from the sets of available options that will best get the AI towards its goal.
Deep Mind is already using this technique in virtual environments – they use computer games for developing and testing out AI algorithms. Virtual environments and computer games are much more efficient to test out the capabilities of AI systems than real-world robotics (slower, messier and more expensive). With computer games, it’s also much easier because you can have a feedback of the progress of your AI systems with information like the score etc. Here is a video showing this new technique in action:
All the AI System gets are the pixels on the screen and the goal of the AI is to maximize the score. Everything else is learned completely from scratch. After hundreds of games, the AI realize that the best strategy is to “dig a tunnel into the bricks”. The beauty of it is that the researchers (who are not so good players at Breakout) learned from the AI.
Another important aspect of research at DeepMind is systems neuroscience which means they get their inspiration not only from mathematics and machine learning but also from how the brain works. Their goal is to solve intelligence and to do that, they want to build a kind of Artificial Hippocampus which is responsible for memory, imagination, attention, navigation, planning etc. Deep Mind is getting closer to human-level AI, once they get there, they will use it to solve problems in the real-world (healthcare, climate change…).
Thanks so much for reading, that’s the end of Article 76. Please, share it and don’t forget to follow me on social media. You can also subscribe to my newsletter below to never miss any article. Don’t hesitate to tell me what you think of this Article in the comments and stay tuned for another one tomorrow!