Samsung Banner sticky Advertisement
  • Home
  • Articles
  • Google's AI designs its chip so quickly it can beat human engineers

Google's AI designs its chip so quickly it can beat human engineers

14th June 2021
"Google's own TPU (tensor processing unit) AI chip will incorporate the new machine learning approach in its next edition"

Google has switched to an artificial intelligence (AI) technique for chip design, saying that it can create silicons faster than a human can. This technology will reduce space and power while enhancing performance, in addition to allowing for speedier chip design.

Google AI chip design: The news was first reported in this week's issue of Nature, which described how graph arrangement was used for chip "floorplanning." For years, Google has been creating processors based on machine intelligence. This time, though, the trial appears to be quite successful.

Google's own TPU (tensor processing unit) AI chip will incorporate the new machine learning approach in its next edition. Similarly, a processor with this design will provide greater AI computation.

Chips that are better than humans

Although there are various techniques and components to the chip manufacturing process, producing a clear floor plan is a difficult task. According to the IT giant, it used 10,000 alternative chip floorplans to train a reinforcement learning system in order to increase efficiency and performance.

The AI was able to design chip "floorplans" better than human engineers could after months of work in just 6 hours.

The authors of the study, lead by Google's co-heads of machine learning for systems, Azalia Mirhoseini and Anna Goldie, noted, "Our method has been utilized in production to create the next version of Google TPU."

The floorplan of a chip is a diagram that shows where components such as CPUs, GPUs, and memory are positioned on the silicon die. Their location on the circuit board is crucial since it affects the chip's power consumption and speed.

This has "huge consequences" for the chip business, according to Google engineers.

In complex games like Go and chess, AI has already proved its capacity to defeat humans. Algorithms are well-trained to make the proper move and win the game while gaming. Similarly, the system understands how to make chip design as computationally efficient as possible.

Also Read:

DISHHOME TO DO THE LIVE COVERAGE OF EURO CUP 2020, WITH COMMENTARY IN NEPALI

HYUNDAI NEPAL FOCUSES ON EV INFRASTRUCTURAL DEVELOPMENT & SERVICE SETUP

MI 11 LITE LAUNCHED IN NEPAL

IOS 15 VS. ANDROID 12

CG NET BECOMES THE INITIATOR OF TREND THAT PROVIDES HIGH SPEED INTERNET IN LOWER PRICES AMONG THE ISPS IN NEPAL

Compiled by : Rahul Shrestha Rahul Shrestha