By Marissa Mitchell, Sophia Lopez and Kylee McDonald
MAY 12, 2017
The tech industry booms with new advances and gadgets every day. One of the most rapidly expanding sectors of technology is Artificial Intelligence (AI). Artificial Intelligence is a field that aims to create machinery that imitates the characteristics of the human brain. The intelligence that AI technology exhibits is a product of Machine Learning. Machine Learning is defined as “the practice of using algorithms to parse data, learn from it, and then make a determination or prediction about something else in the world” (nvidia.com). Deep Learning, the use of neural networks, is a subset of Machine Learning.
Neural networks, as defined in an article by Cade Metz of wired.com, are “complex mathematical systems that can learn tasks on their own by analyzing vast amounts of data.” As companies such as Google, Facebook, and Microsoft use AI chips to run these neural networks, they are providing solutions to problems of inefficiency in the tech world and fulfilling the possibilities dreamed of by sci-fi authors.
Many companies are working to improve the process of training and executing neural networks. Nvidia, a prominent global computer science company, makes the graphic processing units (GPUs) that are typically used to expedite the process of neural network training that must occur before the chips perform the tasks. According to Nvidia’s website, the GPU, invented by the company in 1999, was originally meant to be a force in the world
of computer graphics. Today GPUs are a force in the world of AI as “GPU deep learning” has sparked “a string of ‘superhuman’ achievements in image and speech recognition.”
In recent years, Google built its own AI chip called the Tensor Processing Unit (TPU). After the Google Voice feature was created for Android phones, the company’s engineers feared their computer network was no longer large enough. Google already had several data centers and was also beginning to use deep neural networks to run its voice recognition services. The most energy-conserving and cost efficient solution to the company’s need for more horsepower was to build the TPU. According to CNBC reporter Ari Levy, without this chip it is likely that Google “would have had to double its data centers to support even a limited amount of voice processing.” Thousands of these chips are now packed into the machines of the company’s data centers. The chips automate a wide range of tasks, from voice recognition in Android smartphones to selecting search engine results.
The use of neural networks in Artificial Intelligence is also affecting other industries. Lyrebird, a company started at the University of Montreal, specializes in technologies that can use a one-minute recording of a person’s voice to generate any phrase or design a unique voice. This project, although genius, breeds security concerns. A recently released smartphone app called FaceApp uses artificial intelligence to transform faces by adding smiles, even swapping genders. These apps demonstrate that neural networks can be made to generate their own data based on the data used to train them. Because of this ability, these networks can create images and will likely be also able to generate and manipulate video in the future to create original productions.
MIT Technology Review reported that deep neural network technology has been used to detect skin cancer in images of affected areas. Also a contest was held to search for the best algorithms capable of using neural networks to detect signs of lung cancer. As research progresses, the possibility for applications will be virtually endless in the world of AI.