Luminous Computing Uses Photonic Computing To Speed Up AI

 Artificial intelligence is not just a large business, with projected global spending of nearly $422 billion by 2028, but it also necessitates big data computing on a scale that few can fathom.

The present known capabilities of the current computing environments are being pushed by AI research initiatives from Alphabet’s DeepMindAI, Meta, OpenAI, and others. These algorithms improve by expanding in size. Since few years ago, the amount of computer power needed to train the largest AI model in the world has doubled every three and a half months, considerably exceeding Moore’s law.

That dynamic is being attempted to modify by Luminous Computing. By creating a new supercomputer based on photonics chips, the Mountain View, California-based start-up, founded in 2018 by Marcus Gomez, CEO, Mitchell Nahmias, CTO, and Michael Gao (who has since left the company), intends to push cutting-edge technological advancements to maximise artificial intelligence. The founder’s journey narrative that follows is based on my conversations with Gomez and Nahmais.

“About an hour was needed to train the largest AI models on a single system ten years ago. The largest models of today require tens of thousands of machines and hundreds of engineers, taking months to complete. The issue is that we’ve used up all of the hardware’s potential, according to Gomez. A supercomputer capable of supporting the next wave of AI is what Luminous is working to build.

The founders of luminous set out to overcome the difficulty of delivering orders of magnitude improvements in computing performance without sacrificing usability, programmability, or customer satisfaction in order to carry out their mission of enabling AI to fully realise its potential to automate numerous human activities and enhance life. Communication—between chips, between memory, between boards, between boxes, and between racks—turns out to be the bottleneck. And according to Gomez and his co-founders, photonics-based computer architecture holds the key to the solution.

“I have a background in optics. We are aware that optics offers a solution to the communication issue. For this reason, data is sent across oceans and between racks in a data centre using optics. Therefore, to directly address this issue, we are integrating optics into the computer architecture, explains Nahmais. By redesigning the system with optics to overcome the previously impassable performance-usability curve, Luminous seeks to employ optics to address the data movement problem by providing both performance and ease of use while removing the complexity of the software stack.

Even though production for Luminous, which has been in development for four years, is still 20 months away, Gomez claims that the business already has clients lined up for when it does. However, the concept’s potential as well as the background and advancements that the founders and their expanding team of engineers have made to date have enabled the business to attract the type of venture financing necessary to develop and create a new supercomputer based on photonics.

According to Gomez, the company has so far received $126 million in capital. On March 3, 2022, $105 million was raised in their most recent A round. Bill Gates, Neo, Alumni Partners, Strawberry Creek Ventures, Gigafund, Modern Venture Partners, Horsley Bridge Partners, Third Kind Venture Capital, 8090 Partners, and others are among the investors.

Luminous Computing

Gomez is the son of Indian immigrants who arrived in the country of the United States with less than $1,000. They relocated to a tiny farming community in Minnesota. “I wasn’t really exposed to technology when I was growing up. Pop science magazines were how I first learned about cutting-edge technologies, claims Gomez. Because he wanted to focus on solving fundamental human problems and because artificial intelligence was meant to be the basis on which subsequent discoveries would come from, he was motivated by Steve Jobs’ advice to think big and moved to Silicon Valley to attend Stanford. He earned a computer science degree from Stanford and went on to work in the field of artificial intelligence while pursuing his degree and then for Google. He eventually met Nahmias, and the two collaborated to found Luminous Computing in 2018.

I waited until I was 22 to receive my driver’s licence because I was so confident that self-driving cars would be widely available, claims Gomez. But when he arrived in Silicon Valley, he was dissatisfied and angry because there was still a wide disparity between what he believed was feasible and what was actually taking on there. For over ten years, he has been working on various aspects of artificial intelligence. And when Nahmais showed him what he had been working on, it was clear right away that artificial intelligence would advance significantly if they were able to accomplish even a small portion of it.

Photonic Computing

Although Nahmias was born and raised in Menlo Park, California, in the heart of Silicon Valley, he moved to Princeton University in New Jersey for college in order to study alongside Albert Einstein, the university’s most illustrious speaker and scientist. Nahmais claims, “I came to Princeton to study how to create the next supercomputer. At the time, he believed that developing a method for harnessing physics to create new gadgets to create new technologies would be the answer to creating the next AI supercomputer. For his groundbreaking work in the field of Neuromorphic Photonics, he would later receive his Ph.D. He realised it’s not just physics when he met Gomez. Supply networks, economics, and usability are just as important as physics in this context.

High hopes are being placed in the company, but only time will tell if Luminous can live up to its promise. But Gomez isn’t lacking in big-picture thinking. According to Gomez, “the minimal standard that we’ve set for ourselves is that we want to be the hardware provider that every AI practitioner on Earth utilises.”

Leave a Reply

Your email address will not be published.