AI designed computer chips so complex that humans can’t understand them

Computer chips are everywhere. They power our smartphones, run our cars, and even manage our household appliances. With such intense demand, the race to build … The post AI designed computer chips so complex that humans can’t understand them appeared first on BGR.

Feb 4, 2025 - 11:02
 0
AI designed computer chips so complex that humans can’t understand them

Central Computer Processors CPU concept - 3d rendering

Computer chips are everywhere. They power our smartphones, run our cars, and even manage our household appliances. With such intense demand, the race to build faster, more efficient chips never slows. In an attempt to find new ways forward, some scientists have turned to AI-created computer chips. There's just one problem: they're so complex we can't understand them.

The research has been led by a team of scientists at Princeton University’s Sengupta Lab. The team has been using inverse design to try to create new chips using AI. Their approach doesn’t aim to replace human ingenuity, though. Instead, they want to enhance it.

They want to allow AI to handle the heavy lifting of complex calculations while humans focus on creativity and problem-solving. They used a form of AI called a convolutional neural network (CNN). While human-made chips are tidy and orderly, AI-created computer chips appear more chaotic and unconventional. Humans keep things orderly. But, because AI can process ideas at a much faster pace, it comes up with designs that don't make much sense at first glance.

Artififical IntelligenceImage source: Kaikoro/Adobe

In fact, they’re so different from what humans create that the researchers say it’s hard to even understand their design. However, this apparent messiness could hide a new level of efficiency. Unlike human engineers, AI isn’t bound by traditional design rules, which may allow it to discover patterns and solutions that would take humans years to find.

One of the most exciting aspects of the AI-created computer chip is its potential to redefine what’s possible in chip performance. AI can generate new design paradigms in minutes, offering fresh ideas that human engineers might never consider. This synergy between humans and machines could lead to groundbreaking advances in wireless communication, data processing, and beyond.

However, there are huge hurdles to overcome. AI tends to produce hallucinations—results that seem logical to the algorithm but are impossible to implement in the real world. These hallucinations are most commonly seen in AI like ChatGPT, where the model generates information it deems fit for the requested task.

This is where human oversight becomes crucial. Additionally, the concern that humans may never fully understand these AI-created chips is a huge obstacle that we have to overcome. still, the researchers hope that it will help engineers have their own “aha!” moments and come up with new ways to revolutionize chip designs.

The team published their work in Nature Communications. The hope is that together, more scientists can find a solution that works—or at the very least, that someone can take advantage of the researchers’ CNN technique to come up with new designs that allow for greater efficiency and power.

Combined with other chip advancements—like the DNA chips we’ve heard about recently—these AI-created computer chips could be the start of a very bright future for computers. We just have to figure out how they work first.

The post AI designed computer chips so complex that humans can’t understand them appeared first on BGR.

Today's Top Deals

  1. Today’s deals: Record-low prices on AirPods Pro 2, iPad 10, AirPods 4, Apple Watch Series 10, more
  2. Best Apple deals for February 2025
  3. Today’s deals: $17.50 Apple AirTags, $79 Vizio soundbar, $500 off eufy Omni S1 Pro, $180 Beats Studio Pro, more
  4. Today’s deals: $155 unlocked iPhone SE, $279 iPad 10th-Gen, $24 TP-Link WiFi extender, $279 Galaxy Ring, more

AI designed computer chips so complex that humans can’t understand them originally appeared on BGR.com on Mon, 3 Feb 2025 at 19:51:00 EDT. Please see our terms for use of feeds.