AI chips are designed to perform complex AI tasks efficiently. They consume much less power and provide much more data security, and latency. But designing an AI chip has its challenges. But what if we could train an AI chip to develop itself, pack more circuitry in a fingernail-sized chip while maintaining power, speed, and energy efficiency. Google is proposing exactly that. An AI solution that could advance the internal development of a chip in itself.

The agenda behind this proposal is to shorten the AI chip design cycle creating a correlative relationship between AI and Hardware. Google has developed a family of AI hardware over the years, including the Tensor Processing Unit for processing AI in its servers.

Chip designing includes “floorplanning“, which is a very tedious process even with today’s technology. Hence come to the rescue AI, which studies the existing circuitry to develop its own. It not only takes less time than a human but also analyzes millions of design possibilities as opposed to thousands.

When these AI chips do these tasks repeatedly, they keep updating their parameters with every power reduction, area reduction, and performance improvement, which results in higher performance.

AIMonks #AI #ArtificialIntelligence #AIChip #Google #Hardware #Circuitry #Floorplanning #Technology