“Inexact” Computer Chip Is 15 Times More Efficient Than Perfect Counterparts

Recommended Videos

If you’ve ever done any programming, you know that one of the best things about computers, and also one of the most annoying things about them, is that they don’t make mistakes. A computer will always do exactly what you tell it to, and if you don’t get the right result, it is 100% your fault. At least, this is usually the case. Researchers at Rice University have made an interesting discovery, however. If you build your computer chips so that they can make mistakes, you can boost your efficiency by crazy amounts, to the point where more accident-prone chips can be 15 times more efficient than their always-perfect counterparts.

How do you build a chip that can make mistakes? Simple, you take out some of the components that are rarely used. This gives you increased speed, a smaller size, and more simplicity overall and the only price you have to pay is that the chip can now start making the mistakes that those components where there to prevent. Depending on what you’re doing, this can be an extremely worthwhile trade-off.

Now you might be thinking “but mistakes are bad!” and they usually are. Just the right amount of mistakes, however, can be close to unnoticeable. Above, are three copies of the same image. Leftmost is a copy that was processed with 0% error. The center image has a relative error of 0.54%, the rightmost 7.58%. That last one leaves a little something to be desired, but it’s still recognizable thanks to you eyes’ ability to control for oddities.

For jobs where your eyes, or ears, can perform their own, instictive auto corrections, these chips could come in very handy. The hope is that they can be incorporated in hearing aids and tablets for students in India by 2013. If the use of these kind of chips really catches on, it could mean a whole new tier of electronics that are far more efficient — and cheaper — than other models on the market, but at the cost of more errors in the form of slightly distorted video or audio. Whether or not there’s a market for that kind inaccuracy is something we’ve yet to find out, but at least if computers can make mistakes, maybe we’ll be able to defeat SkyNet after all.

(via The Verge)

I sure could go for some chips


The Mary Sue is supported by our audience. When you purchase through links on our site, we may earn a small affiliate commission. Learn more
related content
Read Article ‘Mamma Mia!’ Star Sara Poyzer Says a BBC Production Replaced Her With AI
Sara Poyzer performs at the Magic at the Musicals event in 2019
Read Article In Moment of Unbelievable Irony, Midjourney Accuses Stability AI of Image Theft
Spider-Man pointing at another Spider-Man, who is pointing back.
Read Article Elon Musk May Be the Lesser of Two Evils in This Legal Battle With OpenAI
Elon Musk at the 2022 Met Gala
Read Article A.I. Scammers Are Impersonating Real Authors to Sell Fake Books
A robotic hand holds a pencil.
Read Article Sexist Trolls Drive Away Twitch’s Top Female Streamer After 10 Years
Imane "Pokimane" Anys at the 2023 Green Carpet Fashion Awards
Related Content
Read Article ‘Mamma Mia!’ Star Sara Poyzer Says a BBC Production Replaced Her With AI
Sara Poyzer performs at the Magic at the Musicals event in 2019
Read Article In Moment of Unbelievable Irony, Midjourney Accuses Stability AI of Image Theft
Spider-Man pointing at another Spider-Man, who is pointing back.
Read Article Elon Musk May Be the Lesser of Two Evils in This Legal Battle With OpenAI
Elon Musk at the 2022 Met Gala
Read Article A.I. Scammers Are Impersonating Real Authors to Sell Fake Books
A robotic hand holds a pencil.
Read Article Sexist Trolls Drive Away Twitch’s Top Female Streamer After 10 Years
Imane "Pokimane" Anys at the 2023 Green Carpet Fashion Awards
Author