What’s the catch?Following two years of supply chain disruptions from the coronavirus pandemic, Russia’s war in Ukraine is another foe for the semiconductor industry. It’s a moonshot and yet, the young company. That’s why Harvard and Boston University partnered with Lightmatter to apply its electro-photonic computing solutions to the AV problem. Luminous Computing, a one-year-old startup, is aiming to build a photonics chip that will handle workloads needed for AI at the speed of light.Since photonics uses less energy than electronics, it’s also being applied to AVs, which have to process large amounts of data from energy-intensive LiDAR, radar, and cameras, straining vehicle batteries and reducing range.Today, models have 10 trillion parameters requiring tens of thousands of machines a year to train-a slowdown that photonics can speed up to just a few days of training. According to Luminous CEO Marcus Gomez, AI models from a decade ago had 100 million parameters at most, taking a single computer an hour to train.It could also affect us closer to home with faster IoT and downloads speeds and higher streaming quality.
With the drive for innovation increasingly bumping up against data bottlenecks, photonics can advance the most demanding areas of technology like AI, quantum computing, and automation. Lightspeed data processing:Since photonic chips use light to transmit data instead of the electrons used in their conventional counterparts, the technology promises faster, more powerful, and more efficient computers. Hewlett Packard Enterprise struck a multi-year strategic collaboration with Ayar Labs to harness the startup’s silicon photonic tech, which can reportedly achieve 1,000x the bandwidth of electrical I/O circuitry with just one-tenth of the power, for use in the HPE Slingshot, per ZDNet.
LUMINOUS COMPUTING AI 105M GATES 200M300MWIGGERSVENTUREBEAT SERIES
0 kommentar(er)
