AI w/ Pi

Interactions between computers and the physical world are forced and awkward. Robots seem preprogrammed and even the smallest of changes in the environment can trip up machines.

In the digital world everything is deliberately coded and predictable. In contrast, the physical world is chaotic and organic and computers have to account for unexpected changes in the environment.  Machine learning provides a bridge into the physical world as it is able to learn from the environment and developers don’t have to account for each and every variable for the algorithm to be successful.

One of the most exciting aspects of machine learning is the ability for a computer to learn and interact with its environment without specific directions. Out of all the fields of ML, reinforcement learning is best suited to this task. Its learning process of interacting with the environment and trying to maximize a specific reward, roughly mimics how we acquire new skills. Reinforcement learning algorithms aren’t guided by their developers to solve problems in any specific way. Instead, they are free to learn and explore their own methods of finding solutions, which often end up being more creative than what a human might have come up with. We have already begun tackling some of the hardest problems using AI, such as self-driving cars and robotic. This is just the tip of the iceberg in terms of what is possible in the near future.

I am going to begin exploring how Reinforcement learning can enable machines to live and learn in the same world as us. I have installed Tensor Flow on a Raspberry Pi thanks to a port by samjabrahams. I also started teaching myself the basics of reinforcement learning and ideating creative applications for it with the Raspberry Pi. In the next few weeks, I will post basic sketches of ideas and various bits of code showcasing what I have learned. My aim is that these ideas/sketches/prototypes will lead to a project that will push the boundaries of reinforcement learning in the physical world.

Stay tuned!