Ricursive Intelligence founders Anna Goldie and Azalia Mirhoseini believe more efficient AI chip designs could help solve the computing crisis. The founders took to the stage at TechCrunch Disrupt on Wednesday to detail how the company is developing an AI that can design chips, learn from each iteration, and improve subsequent designs, all with the goal of automating and speeding up the chip design process.
“There’s a recognition that over the next five years, the demand for computing will increase at least 10,000 times, so this may be a bit conservative,” Goldie said on stage. “So we don’t want to build 10,000 times more data centers. We think we need to help by making chips that are more efficient, use the bandwidth of those data centers more efficiently, make chips that use wafers and fabs more efficiently. Because even if we built (data centers) in space or something, I don’t know, we wouldn’t be able to see the sun anymore.”
While the idea of AI designing its own chips may be frightening to some, Goldie argues that allowing AI to design chips allows humans to focus on developing applications and exploring new possibilities that extend what computing can accomplish.
“I think what’s changing everything is that humans bring in great ideas and creative thinking, and the execution is handled by AI,” Goldie said. “And this is happening in chip design, software development, and other places. And it’s something that happens recursively in chip design.”
The founders said their AI is challenging traditional chip design methods by discovering unconventional layouts that can outperform traditional chip design methods. For example, the founders’ previous AI system, AlphaChip, which had chip components arranged in a circular “doughnut” pattern, initially surprised engineers but ultimately proved successful.
They explained that they are optimizing the entire chip, rather than following the traditional approach of dividing the chip into dozens of small blocks. This approach allows the AI to create complex patterns, including wave-like designs, which the pair said they plan to share more about in the future.
The founders also claimed that AI-designed chips would enable the creation of all kinds of chips that were previously impossible.
“If you have six months, a year, two years to design a new chip, you have to be very selective about the chips you design,” Goldie says. “But if you can design a chip in a week or so and then use it for manufacturing, you can design custom chips for all kinds of applications, including space, AR, VR, and biological computing. It makes economic sense to do so, and it’s also energy efficient.”
If you make a purchase through links in our articles, we may earn a small commission. This does not affect editorial independence.
