Situs Panduan dan Solusi Terkini

Latent AI, which says it can crush popular AI models by 10x, gives some major boost – TechCrunch

  • Share
banner 468x60

About a year ago Latent AI, a startup in Menlo Park, California, that’s now three years old, wooed a few investors during TechCrunch’s Battlefield competition. She did not triumph this competition, but that did not quit her from gaining the attention of investors elsewhere. Just closed on $19 million Series A funding in a round co-led by Future Ventures and Blackhorn Ventures, with participation from Booz Allen, Lockheed Martin, 40 North Ventures and Autotech Ventures. The company has now raised a whole of $22.5 million.

banner 336x280

What exactly are these backers funded? The company says its software tools allow AI models to run anywhere, regardless of hardware limitations, and this includes the cheap chips typically found in lofty-end hardware. It also says it can crush common AI models ten times without noticeable change in fidelity, partly through an “attention mechanism” that enables it to preserve energy and run only what is needed, and partly because it can self-adjust workload based on the environment and operational context .

This means – according to Latent AI – that its tools can aid developers deliver AI models that are optimized for computing; that they can overcome memory and energy limitations; And that there is almost no response time (hence the name of the company).

“Think of confront detection algorithms that work locally inside cameras or security devices, or Siri-like voice interfaces that work instantly, even when there is no network connection,” says Steve Jurvetson, a veteran investor and co-founder of Future Ventures.

Sure, there is a vast market for the type of technology that Latent AI is developing. In fact, although in an interview earlier today, co-founder and CEO Jags Kandasamy declined to go into detail, he suggested that the US government is indeed a client, thanks in part to strategic investors like Booz Allen. (In a press release final month about Booz Allen’s investment in Latent AI, Steve Escaravage, senior vice president at the Government Services Agency, noted that “the aptitude to collect, analyze, and act quickly on data is the core” of the United States’ national defense strategy.)

Another strategic investor, Lockheed Martin, is focused heavily on finding ways that AI technologies can aid the US military improve condition awareness across the land, sea, air, space, cyber and electromagnetic spectrum, so its appeal to a startup is simple to appreciate.

Kandasamy also spoke this afternoon about an unnamed ski equipment manufacturer that is using Latent AI technology in Google Glass-like augmented reality goggles, and suggested that Latent AI sees a world of opportunity in the commercial market as well.

Of course, since the world is now packed of data-collecting hardware and there is a vast interest in making that data actionable without having to send it back and forth to a distant cloud, there are many other companies and projects out there that are just as objective as the underlying AI. Among them are begin source tools such as TensorFlow, hardware vendors such as Xilinx, and competing startups such as OctoML and Deeplite.

Kandasamy insists that everyone falls short in some way. About TensorFlow, he says that developers only have community support when it comes to production deployment. Among the world’s chip makers, they focus on their own hardware and not vertical integration. What about these competitors? Kandasamy says they focus on either compression or assembly, not both.

Either way, Latent AI has the considerate of backstory that investors love. After Kandasamy, a serial entrepreneur, sold his latest startup to Analog Devices, he headed to SRI International in 2018 as a resident entrepreneur. There, he soon became fascinated by technology developed by Sek Chai, who was the research firm’s technical director for nearly a decade, specializing in lofty-performance low-power computing, computer vision, and machine learning. In fact, soon following, Kandasamy had formed Chai’s latent AI and began to clear some kinks.

Now, with unused funding and a little but growing customer base pushing Latent AI on a subscription basis to access its tools (and then deploy them to the workplace), the interrogate is whether the 15-person outfit will come far enough, and rapid enough, to conquer current and coming competitors.

Clearly, Jurvetson thinks she’s in pleasing shape. He says he has been on the advisory board of SRI International for more than a decade and has seen many pleasing technologies like Siri evolve and then exit the organization.

“This is the only one I invested in,” he says.


banner 336x280
banner 120x600
  • Share