Machine learning has come a long way in recent years. Facebook’s Artificial Intelligence Research Lab (FAIR) has developed a number of machine learning tools that go beyond object recognition, and has given them to the public. Developers are encouraged to play with and code the company’s open sourced tools. Although FAIR is not the first company to release its tools, one thing that sets it apart is that the tools recognize two-dimensional images rather than just depth sensors. Also, rather than putting a box around images, it can almost see each detail in the photo.
Facebook hasn’t made use of these tools just yet, FAIR released them early, allowing developers to work on them and solve short-term problems that can be built upon and into a bigger picture for future Facebook projects.
Tech giant Google is also working diligently to develop several AI programs. Google’s TensorFlow machine using technology is used to save farmers’ time with sorting cucumbers, using machine learning to recognize each cucumbers’ attributes such as color, shape and size. An Arduino Micro uses the information to control the actual sorting and a Windows PC trains the neural network with images. Currently, it takes two or three days to train the program using low-resolution photos, but this example shows just how advanced and profitable machine learning can be.
Machine learning can be beneficial for the environment. One of the major obstacles that big data centers face is keeping their servers cool because of the amount of energy used. Google is now using its DeepMind artificial intelligence to manage the power usage in parts of its data centers. The company tracked variables like temperature and pump speed to come up with an energy-efficient cooling method and ended up reducing the amount of electricity needed for cooling by 40 percent.