AI is advancing at a rapid rate and with such a speed the industry is demanding more innovation to speed up the computing process while keeping power consumption to a minimum. Conventional semiconductors and AI chips aren’t ideal for performing tasks that require high computational power as the increasing energy consumption makes it almost implausible.
A startup called Luminous Computing wants to overcome this hardware bottleneck and boost the performance of semiconductor chips. Their answer to this bottleneck is light. Compared to currently deployed semiconductors that use electrons to perform the computational process for training AI models, Luminous Computing wants to use lasers to beam light through tiny structures on its chips.
How Does Light Help?
These tiny structures, known as waveguides, guide light waves with minimal loss of energy. The startup aims to use different colors of light to transfer bits of data through waveguides simultaneously. The process can increase the efficiency of electrical chips, making them more efficient in data-carrying than conventional electrical chips.
Since a large amount of data can be transported at the same time, these processors are best suited for AI processes that require high computation.
Moreover, optical processors require less power as compared to electron-based processors.
Backed by Bill Gates, Uber CEO
Luminous Computing is backed by many popular names in the business and tech industry, including Microsoft founder Bill Gates and Uber CEO Dara Khosrowshahi. The startup has recently received $9 million in seed funding from investors.
Luminous’ co-founder Mitchell Nahmias says that the current prototype based on the optical processors is three times as energy efficient as the top of the class conventional AI chips. You can see Luminous’ optical processor in the featured image at the top.
There Are Hurdles
According to Dirk Englund, an MIT professor who also holds the position of a technical advisor at Lightmatter says that it will be challenging for Luminous to manage device required to manipulate light and use it for data transfer. To control light, the startup would need a host of devices like electro-optic modulators and lasers.
Luminous Computing isn’t the only company trying to boost AI chips using light. There are other startups like Lightelligence and Lightmatter that are also working on optical chips. Prominent companies like Intel are also conducting research in the field and are looking forward to optical processors.
Investors including Gates and others trusting the company that it will overcome the hurdles to develop processors that will help in accelerating AI.