Recently, Luminous Computing raised $9 million in seed funding from well-known investors including Bill Gates and Uber CEO Dara Khosrowshahi. Luminous Computing has an ambitious plan to accelerate artificial intelligence with a new chip: traditional semiconductors use electronics to help with the mathematical calculations needed for artificial intelligence models, and Luminous uses light.

Many industries, including autonomous vehicles and drone manufacturers, are trying to implant more and more artificial intelligence into their machines. But electronic chips that are widely used like CPUs are not suitable for these tasks because they consume a lot of power and are not fast enough to process data.

These limitations can lead to lags and delays—such as waiting for machine learning results to write a thesis is annoying; but if you rely on artificial intelligence algorithms to guide a car on a busy street, the consequences will be much more serious.

At present, the bottleneck of artificial intelligence is getting more and more serious. A study by the artificial intelligence research firm OpenAI shows that the computing power required to train the largest artificial intelligence models doubles every three and a half months.

Marcus Gomez, Luminous's CEO and co-founder, points out that while there are a lot of hype around artificial intelligence, the limitations of the underlying hardware are hampering progress. He said: "Several years ago, Silicon Valley promised us that it will launch the artificial intelligence-driven version of Star Trek, but we are still waiting for it. More powerful artificial intelligence chips can push everything from A machine learning model that helps doctors make medical diagnoses, to new artificial intelligence-driven applications that can run on smartphones.” Luminous believes that light is the solution to this problem. It uses a laser to emit light through tiny structures on the chip called waveguides. By using different colors of light while using a waveguide to move multiple blocks of data, it can surpass the data carrying capacity of traditional electronic chips.

The ability to quickly transfer large amounts of information indicates that optical processors are well suited for processing large amounts of computations that drive artificial intelligence models. At the same time, they consume much less energy than electronic chips.

Mitchell Nahmias, another co-founder and chief technology officer at Luminous, said the company's current prototypes are three orders of magnitude more energy efficient than other state-of-the-art artificial intelligence chips. The startup's processor was developed based on years of research by Nahmias and other scholars at Princeton University.

However, Luminous still faces fierce competition. Startups like Lightelligence and lightmatter are also working on developing optical chips to accelerate artificial intelligence. In addition, semiconductor giants like Intel are accelerating research in the field and may introduce new optical processors.

Dirk Englund, a professor at the Massachusetts Institute of Technology and a technical consultant at Lightmatter, believes that Luminous may face challenges in mass production. For example, optical chips need to control light from lasers to electro-optic modulators to make them work, which is an important reason why optical chips have not been widely used.

Gates and other sponsors bet that Gomez, Nahmias and Luminous's other co-founder Michael Gao can overcome obstacles. They also bet that companies that break the computational bottleneck will unlock the true potential of artificial intelligence.

Ali Partovi, who invested in the Luminous Seed Fund's Neo Venture Fund, points out that even things like voice assistants on smartphones can easily fail because they lack sufficient artificial intelligence computing power.