Samsung uses artificial intelligence to automate the insanely complex and subtle process of designing advanced computer chips.
The South Korean giant is one of the first chipmakers to use AI to make its chips. Samsung is using AI features in new software from Synopsys, a leading chip design software company used by many companies. “What you see here is the first of a true commercial processor design with AI,” said Aart de Geus, Synopsys president and co-CEO.
Others, including Google and Nvidia, have talked about designing chips with AI. But Synopsys’ tool, called DSO.ai, is perhaps the most far-reaching because Synopsys works with dozens of companies. The tool has the potential to accelerate semiconductor development and unlock new chip designs, according to industry watchers.
Synopsys has another valuable asset for creating AI-engineered chips: years of advanced semiconductor designs that can be used to train an AI algorithm.
A Samsung spokesperson confirmed that the company uses Synopsys AI software to design its Exynos chips, which are used in smartphones, including its own branded handsets, as well as other gadgets. Earlier this week, Samsung unveiled its latest smartphone, a foldable device called the Galaxy Z Fold3. The company has not confirmed whether the AI-designed chips have already entered production or which products they may appear in.
Across the industry, AI appears to be changing the way chips are made.
A Google research paper published in June described using AI to rank the components on the Tensor chips it uses to train and run AI programs in its data centers. Google’s next smartphone, the Pixel 6, will feature a chip manufactured by Samsung. A Google spokesperson declined to say whether AI helped design the smartphone chip.
Chipmakers, including Nvidia and IBM, are also working on AI-driven chip design. Other chip design software makers, including Cadence, a competitor to Synopsys, are also developing AI tools to map out the blueprints for a new chip.
Mike Demler, a senior analyst at the Linley Group who tracks chip design software, says artificial intelligence is well-suited to arranging billions of transistors on a chip. “It lends itself to these issues that have become immensely complex,” he says. “It’s just going to be a standard part of the calculation toolkit.”
Using AI is often expensive, Demler says, because it requires a lot of cloud computing power to train a powerful algorithm. But he expects it to become more accessible as computing costs fall and models become more efficient. He adds that many tasks associated with chip design cannot be automated, so expert designers are still needed.
Modern microprocessors are incredibly complex and contain multiple components that must be effectively combined. Sketching out a new chip design normally requires weeks of painstaking effort and decades of experience. The best chip designers use an instinctive understanding of how different decisions will affect each step of the design process. That understanding cannot be easily written in computer code, but some of the same skill can be captured using machine learning.
The AI approach used by Synopsys, as well as Google, Nvidia and IBM, uses a machine learning technique called reinforcement learning to flesh out the design of a chip. Reinforcement learning involves training an algorithm to perform a task through reward or punishment, and it has proven to be an effective way to capture subtle and hard-to-codify human judgment.
The method can automatically establish the basics of a design, including the placement of components and how to connect them, by trying different designs in simulation and learning which produce the best results. This can speed up the process of designing a chip and allow an engineer to experiment with new designs more efficiently. In a June blog post, Synopsys said a North American integrated circuit maker had improved a chip’s performance by 15 percent using the software.
Most famously, reinforcement learning was used in 2016 by DeepMind, a subsidiary of Google, to develop AlphaGo, a program that can master the board game Go well enough to beat a world-class Go player.
De Geus says his company realized that reinforcement learning could also be useful for chip design. “Just over a year and a half ago, we were able to achieve the same results for the first time as a team of experts would get within a few weeks in several months,” says de Geus. He will present details of the technology and its development at HotChips, a conference on semiconductor technology, on August 23.
Stelios Diamantidis, senior director of artificial intelligence solutions at Synopsys, says the DSO.ai software can be configured to prioritize different goals, such as performance or energy efficiency.
Semiconductors, as well as the tools used to make them, have become increasingly valued assets. The US government has tried to limit the supply of chip technology to China, a key rival, and some politicians have called for software to be added to the list of export controls.
The emerging era of AI-designed chips also raises the prospect of using AI simultaneously to modify software to work more efficiently on a chip. These could be the neural network algorithms that run on specialized AI chips and are commonly used in modern AI.
“AI-powered software and hardware co-design is a fast-growing direction,” said Song Han, an MIT professor who specializes in AI chip design. “We’ve seen promising results.”
This story originally appeared on wired.com.