DeepMind Hassabis Now Supports Graphcore AI Chips

Demis Hassabis the Google DeepMind Founder and Zoubin Ghahramani the Uber’s Chief Scientist, are few investors in Graphcore that make chips and process machine learning more efficiently than Graphic processing units(GPUs) and central processing units (CPUs). Graphcore chips are known as intelligent processing units (IPUs). They are designed to create highly intelligent computer systems that can learn for themselves when fed large amounts of data. The Bristol based machine learning startup, Graphcore, has raised $USD30 million to ship its AI based computer chips. The all over investment is now $USD62 million. A prediction has been made that by 2020, AI technology will be everywhere.

Graphocore’s main aim is to become a leader in the machine intelligent processor world, therefore it builds processing hardware and software framework. Earlier, Graphcore was said to be the most disruptive startups in the whole world, now it is all set to spend the funds on its technology as well as its business. Demis Hassabis created a system which is capable of general artificial intelligence. The algorithms are made to learn from the raw data and then use that for various tasks. Graphcore system has become very important to make technologies powerful. In the near future, it will be used for AI based cancer detectors. Having looked at the development in the AI field, we can say that it will take around 20 years for AI innovation to take place.

Also Read: Microsoft Releases Second Update of Its Deep Learning Toolkit

The Atomico partner, Siraj Khaliq, will soon join the Graphcore board of directors. He believes that the performance of software depends on hardware and when it comes to AI, the recent hardware is not upto the mark. He has the deep knowledge about the machine intelligence. The major difference between the design of IPU vs GPUs that the latter is more rigid and connected together at the same time. On the other hand, thousands of processors do machine learning tasks separately. While for IPUs we have to get the processors work together and create an easy environment for the processors.