- Cerebras Systems is an American technological firm based in Sunnyvale, California that works primarily in the semiconductor industry and builds complex Artificial Intelligence(AI) deep-learning applications.
- They are credited with building the world’s largest chip as a part of their CS-2 system which can replace more than tens to hundreds of conventional graphics processing units(GPU).
- CEO Andrew Feldman believes that his company is building effective systems which weigh far more in terms of capacity and performance than in terms of its disputed size.
In a world moving swiftly towards the minimalistic approach in every aspect, whether it be in the field of architecture, design, or technology, it seems as if some people still believe staunchly to live by the quote “bigger is better”. In an era where innovators of the likes of Elon Musk and his scientists at Neuralink are planning to implant coin size brain implants via robotic surgery to integrate human beings with the world of Artificial Intelligence(AI), on the other hand, employees at Cerebras Systems have gone a few steps and inches ahead.
What is Cerebras Systems?
Cerebras Systems is an American Artificial Intelligence(AI) company with its headquarters based out of Sunnyvale in California. It was founded in 2015 and is headed by Andrew Feldman as CEO. The founders were Feldman, Lauterbach, James, Lie, and Fricker all of whom used to work together at SeaMicro, a firm founded by Feldman and Lauterbach in 2007. The firm has its offices based in San Diego, Toronto. and Tokyo. They work primarily in the semiconductor industry and build complex Artificial Intelligence(AI) deep-learning applications.
What are deep learning systems?
To understand the chips Cerebras Systems make and their applications let’s understand what is meant by deep learning. Deep learning is a class of machine learning algorithms that uses multiple layers to progressively extract higher-level features from the raw input. In simpler terms, deep learning models are based on artificial neural networks known as convolutional networks, used to analyze visual imagery. The term “deep” is used to refer to the use of multiple layers in the network.
Artificial Neural Networks are integrations between information processing and biological systems. Although these have many variations with biological information processing which occurs in the brain, in terms of being static rather than dynamic, these networks are the basic structure of a deep learning architecture. Deep-learning architectures have been applied to fields including natural language processing, computer vision, speech recognition, bioinformatics and drug design.
The world’s largest chip
Cerebras Systems has built the world’s largest chip. Many experts have hilariously compared the size to that of a pizza or a big dinner plate. These humongous chips come in the era of various technology giants trying to make sleeker and in general smaller parts that are big in terms of utility. When compared, the Cerebral chips are 50 times larger than the biggest Graphics Processing Unit(GPU) on the market.
According to the technology firm’s website, their CS-2 system can look to replace an entire rack of processors used typically to train AI models. A single CS-2 system delivers the performance of many tens to hundreds of graphic processing units(GPU). This CS-2 system delivers answers in minutes or hours that generally would take days or even longer in conventional multi-rack cluster processors. Hence this technological marvel not only offers massive performance advantages but also effectively reduces the amount of power consumed during AI training sessions.
Are these bigger chips really worth their price tag?
While experts have been going back and forth regarding the actual utility of Cerebras-made chips, CEO Andrew Feldman believes in the beauty of the technology they are creating. Feldman exclaims that his team is composed of hard-working professionals who are building a future that definitely promises a spot for the ambitious company in the Computer Hall of Fame. When asked about the heat which his firm faces amongst the other rising giants of the Silicon Valley, Feldman believes there’s no competition at all. He is confident that the solutions his firm offers with respect to capacity and performance are worth cutting the race for. However, the real advantage of such a large chip versus its ability to replace multiple smaller and easy to be operated components is still debatable
