A computer that uses light instead of digital switch for calculation can help reduce energy requirements Artificial intelligence (AI), according to a new research. Scientists who invented a computer have called it a new computing model.
Microsoft researchers developed a prototype analog optical computer (AOC) that can do something for which AI can be used, as well as correction issues.
Scientists, published in the Journal on September 3, wrote, the new computing system can one day solve certain problems and is worth it with less energy than modern digital computers. Nature.
The co -author of the study’s co -author of the study has provided AOC. Jeans glandroAn AI researcher of Microsoft, said in a Microsoft Blog Post. “It’s never heard in hardware.”
In new research, Microsoft researchers have created a computer that uses micro -leads and camera sensors to calculate.
Unlike regular digital computers, which transfers billions of small switches to calculate, the new system uses different intensity light and voltage to add numbers to the loop and multiply. The OOC counts a problem several times, each time improves the last, even as it reaches a “stable state” or a final solution.
Related: The study says ‘Quantum AI’ algorithm already pushes high -speed supercomputers
Since the AOC does not convert the analog signal into digital during calculation, the AOC saves energy and controls some limits of digital computing.
This special method of calculation makes it “not a common purpose computer, for certain AI and correction issues” make it a particular purpose “a stable state.” “ Audogan OzkinUCLA’s optical computing researcher who was not involved in research, told Science directly in an email.
But for these specific purposes, the AOC can offer significant improvement than digital computing, researchers wrote in the study.
A new, light -based computing sample
The team also created a “digital twin” program – a computer model that imitates physical AOC count. This digital twin can be screened to handle more variables and more complex calculations.
“Digital twins are the place where we can work on much bigger problems than it can do to deal with the device itself.” Michael HensenThe senior director of biochemical signal processing at Microsoft Health Future said in a blog post.
The team first had some easy to learn machine learning, such as photos classification. Physical AOC performed a digital computer as well as a digital computer. The team writes that in the future, the big AOC, which can handle more variables, can quickly eliminate digital computer in energy saving.
Subsequently, researchers used AOC digital twins to re-create a 320-BY-320-Pixel Brain Scan image using only 62.5 % of the original data. Digital twins re -reproduced the scan – scientists say MRI could lead to less times.
Finally, the team used the AOC to solve a series of financial problems, which includes finding the most effective way to exchange funds between multiple groups, while minimizing the risk – a challenge that is facing clearing houses daily – with more success than current quantum computers.
For now, the AOC is a prototype. But since future models increase more microids, machines can become more powerful with millions or billions of variables at a time.
“Our goal, our long -term vision is that it is an important part of the future of computing,” HUSHILYA researcher from Microsoft’s Cloud System Future Team said in a blog post.
