Synopsis
Andrew Dai has launched Elorian, a startup aiming to improve AI’s ability to understand and reason about visual information. Backed by major investors, the company is developing specialized models to support real-world applications like robotics and design, addressing limitations in current AI systems despite strong industry funding.So he’s cofounding a research startup to bring an adult into the proverbial room. The company, Elorian, is building AI that can better understand the imagery in the world around it. The breakthroughs could speed the deployment of AI in industries such as architecture, the automotive market and robotics, Dai said.
Elorian is emerging from stealth mode on Thursday with $55 million in funding and a $300 million valuation. The Palo Alto, California-based startup, which has hired more than a dozen people, is working on building better-reasoning visual models. Investors include Striker Venture Partners, Menlo Ventures and Altimeter Capital, with participation from Nvidia Corp. and prominent AI researcher Jeff Dean.
The round was raised in two tranches, with the first capital coming in at a $120 million valuation and the second portion raised at a $300 million level, according to a spokesperson for Elorian. The Information previously reported on some details of the company’s fundraising efforts.
Though the startup isn’t yet generating revenue, it is having discussions with potential customers and plans to release its first publicly available reasoning model in approximately the next 12 months, Dai said. Elorian’s other cofounders are Yinfei Yang, who worked on AI research efforts at both Google and Apple, and Seth Neel, a former Harvard professor who researched data and artificial intelligence.
Elorian’s pitch: In order to help AI understand the world around it, the tech industry needs to build models designed specifically for that purpose. Despite the billions of dollars that companies such as Google and OpenAI have poured into building massive AI models, the tools still do a poor job of, say, visually analyzing satellite imagery or answering questions about what an image is missing but should have, Dai said.
While coding has brought reasoning capabilities to a different level, Dai believes that models still can’t handle major decisions on how to make lighter cars or more efficient rockets.
“These are not things that you can just express with code and have a faster rocket,” he said. “You actually need to design the physical thing – and that design lives in the physical world.”
OpenAI recently pulled the plug on Sora, its popular video generation app, which sparked some concern in the tech sector about the commercial viability of AI video. But Dai stressed that the company was less focused on creating media and more interested in reasoning capabilities.
“It’s much easier to generate something that looks great, but it’s a lot harder to reason about it, to understand it, to explain what it does to someone,” he said.
Elorian is developing its early AI products on top of open source models, which can be freely used and modified. The company is considering releasing smaller versions of its models to the open source community, but will likely keep its flagship version proprietary, Dai said.
Dai is part of a wave of researchers who left labs in pursuit of more specialised work within the ever-changing world of artificial intelligence. For example, You.com founder Richard Socher has been raising capital at a $4 billion valuation for a more advanced AI system — technology that’s still under wraps.
And Periodic Labs, co-founded by former researchers at Google and OpenAI, is currently fundraising at a $7 billion valuation. It launched with a valuation of more than $1 billion.
Dai said he took a different path with Elorian and turned down a higher valuation at inception — partly to ensure that his early employees would reap a meaningful financial gain.
“If you start with a very high valuation, then for them, it’s very hard for them to see a 50x or 100x increase in the value of their equity, and we wanted to give that opportunity,” he said. “Because we know that the co-founders are not going to build the whole company ourselves.”
Brian Zhan, a partner at Striker, said that Dai’s recruiting philosophy — along with his ability to recruit several other top researchers from DeepMind — was a large part of his firm’s interest in backing Elorian.
The company, Zhan expects, will be more efficient than other cutting-edge labs due to his experience building Gemini, which is Google’s artificial intelligence model.
“Everyone else at every other frontier lab is kind of just raising money to do experiments,” Zhan said. “Andrew knows the Gemini recipe — he’s not wasting a single dollar.”
