Google claims AI to design chips is much faster than humans

Recently, Google announced that it has developed artificial intelligence (AI) software that can design semiconductor chips much faster than humans.

Technology giant Google said in an article published in the world’s leading scientific journal Nature (USA) on June 9 that, a semiconductor chip that takes people months to design, the software Their new AI does it in less than 6 hours.

Google claims to use AI to design chips faster than humans

Google said that AI was used to develop the next version of Google’s Tensor Processing Unit (TPU) chip, which is used to run AI-related tasks.

“Our method was used in the manufacturing process to design the next-generation TPU,” said Google’s co-heads of machine learning, Azalia Mirhoseini and Anna Goldie.

In other words, Google is using AI to design semiconductor chips that can be used to create even more sophisticated AI systems.

Specifically, Google’s new AI software can draw a “floor plan” of a semiconductor chip. This basically involves plotting where components such as the central processing unit (CPU), graphics processing unit (GPU) and memory placed on a silicon die correlate, the position of the them on these small boards is important because it affects the power consumption and processing speed of the semiconductor chip.

It takes humans months to optimally design these “floor plans,” but Google’s deep learning system, an algorithm used to perform certain tasks, can reduce Maximum design time.

Similar systems can also beat humans at complex games like Go and chess. In these situations, the algorithms are trained to move the pieces to increase the chances of winning.

However, in the semiconductor chip scenario, the AI ​​is trained to find the best combination of components integrated in it. The AI ​​system is provided with 10,000 “ground plans” of semiconductor chips for them to choose the most optimal implementation.

While semiconductor chip designers often arrange components in neat lines, Google’s AI uses a more distributed approach to its semiconductor chip design.

This is not the first time an AI system has learned to perform a task based on human data. Previously, the popular Go computer program “AlphaGo” by artificial intelligence company DeepMind made a unique move against world Go champion Lee Sedol in 2016 that made Go players on the whole world is amazed.

Google engineers noted in the paper that the breakthrough could have “major implications” for the semiconductor sector.

The breakthrough is hailed as a “major achievement” that will greatly help in accelerating the semiconductor supply chain.

However, Nature also argues that technological achievements must be shared widely to ensure that the ‘ecosystem’ of companies becomes truly global, and that the industry must ensure that the best techniques are available. Time-saving techniques do not lose talent in the field.

Phan Van Hoa (according to CNBC)

Intellectual property should be the foundation of Make in Vietnam products

Intellectual property should be the foundation of Make in Vietnam products

Protecting intellectual property rights and owning patents helps Vietnamese businesses to stand on an equal footing with businesses in the region.


Leave a Reply

Your email address will not be published. Required fields are marked *