 |
| Google's AI designs chips faster than humans |
According to Google, it has developed artificial intelligence software that can develop computer chips faster than humans.
Google is using machine learning to create the next generation of machine learning chips. The company's engineers said these algorithm designs are comparable to or better than those created by humans and can be created more quickly.
The research giant said in a research paper: The development of this chip may take several months, but a new artificial intelligence can be developed in less than six hours.
The company said that artificial intelligence was used to create the latest version of the chip for Google's tensor processing unit.
“Our production method is used to develop the next generation of CPU tensor chips,” wrote the author of the article, under the guidance of the head of machine learning at Google Systems.
In other words, Google uses AI to develop chips that can be used to create more complex AI systems.
Specifically, the new system can set Google Slides.
This basically involves placing components such as the processor, GPU, and memory relative to each other on a silicon mold.
Placing these components on these small boards is important because it affects chip power consumption and processing speed.
Google artificial intelligence:
It takes months to improve these models. But Google's deep reinforcement learning system can easily do that.
Similar systems can defeat people in complex games such as go and chess. In these scenarios, the algorithm is trained to move the chess pieces in order to increase their chances of winning.
However, in a chip scenario, the system can be trained to find the best combination of components to maximize computational efficiency.
The system received 10,000 smart cards to see which are valid and which are not.
Human chip designers usually present components in sharp lines, while Google's AI uses a more decentralized approach when designing their chips.
It is reported that this is not the first time that artificial intelligence has succeeded after learning to perform tasks after entering human data.
In 2016, the popular AI AlphaGo software from DeepMind took an unorthodox step towards becoming Go World Champion, shocking Go players all over the world.
Google engineers pointed out in the paper that this hacking attack could have a significant impact on the semiconductor industry.
Facebook's chief AI scientist praised Twitter's research as "extremely cool work," adding that it was the culmination of improved AI learning.
This breakthrough has been hailed as a major achievement and will go a long way in accelerating the supply chain.