X
    Categories: Featured

Why Google’s Tensor Processing Unit is really important

Image: Seeking Alpha.

On April 5, the Google Cloud Platform blog published an article talking about the Tensor Processing Unit, a new kind of chip the company designed for heavy machine learning operations. The TPU is basically the first version of the Terminator’s brain, and even though it sounds terrible, we just need it.

Let’s begin by labeling the new chip as specialized hardware and understanding it’s not the first of its kind. Most people should know what a CPU (Central Processing Unit) is, in general terms, a computer’s brain. However, machines think differently and need more than one organ dedicated to processing information.

For example, a decent CPU can deal with graphic workloads, but if someone wants to play a modern video game, he or she couldn’t do it without a GPU (Graphic Processing Unit). They usually come in graphic cards and are much more efficient than a CPU when it comes to handling video games, and designing and rendering, among other things.

Is machine learning that important? 

Yes, it is important and incredibly weird. The term refers to a series of process scientists use to teach software a skill, and they have succeeded in the past. Take DeepMind as an example.

The programmers “taught” the AI how to play Go, an ancient Chinese strategy game which is considered to be extremely difficult. Not only did DeepMind learn how to play it but it also beat the Lee Sedol, the then World Champion. It was the first time a computer bested a human Go master.

Moreover, the average Internet surfer benefits from machine learning every day because Google has been improving most of its services with it for years. Some of the programs that use the new technology are Google Translate, Google image, Google Photos, and Google Visions API, among many others.

The future of our society depends on things like the TPU

Artificial intelligence is here to stay, whether people like it or not. It is a fact that AIs can do things better than humans, and ever since companies realized that, more and more workers are being replaced by “trained” software.

The AI ambassadors are self-driving cars, I would say. They are popular, and they both fascinate and terrify people. But, AIs already control massive manufacturing chains around the world. Taking into account the technology is also blended with a lot of modern apps, we need it to be as perfect as it can be.

For instance, many companies in the U.S. are starting to use chatbots to filter people who apply to positions withing their organizations. Its name is Mya, and it compares the applicant’s profile to the firm’s database to determine whether he/she is suitable to the position.

Some people hate the idea of being evaluated by lines of coding while others are excited about the improvements it will bring, but both arguments are irrelevant. The fact is businesses can and are using it, and it will probably become popular in the highest spectrum of the job market.

If that is the case, we need machine learning to be perfect so that AIs do a great job which makes Google’s Tensor Processing Unit maybe the most important development of our generation in the area, of course.

Source: Google Cloud Platform Blog

Hector Morales:
Related Post