The startup, one of the few female-led AI unicorns, is valued at $1 billion and has access to 10,000 Nvidia H100 GPUs, but its founder says it could be years away from unveiling a product.
through Alex KonradForbes staff and Kenrick KaForbes staff
iIt was at a San Francisco party hosted by Greg Brockman nearly a decade ago — when the Stripe CTO was about to leave to co-found an AI research lab called OpenAI — when entrepreneurs Kanjun Qiu and Josh Albrecht met a crypto mogul named Jed McCaleb.
As Qiu and Albrecht launched two now-defunct startups, and McCaleb became a regular at The Archive, their group home for founders and AI researchers near Dolores Park, they periodically talked about starting their own AI lab. Brockman and Sam Altman’s OpenAI project was joined by many of his peers. Still, it felt too early, “a little crazy,” to walk away from what they had built, Qiu said. But when his other startup, Sources, stopped growing fast enough, he reconsidered. And when the duo raised $20 million last year for a new AI research company, Generali Intelligent, McCaleb wrote the biggest check.
Now, Qiu and Albrecht are doubling down on their lab under a new name, Imbue — and this time, McCaleb, the billionaire co-founder of crypto startup Ripple, is writing another big check. Imbue has announced that Astera Institute, its nonprofit organization dedicated to supporting science and technology projects, is leading a $200 million Series B funding round in the startup. The all-cash round, which included AI chipmaking heavyweight Nvidia, Cruise cofounder Kyle Vogt and Notion cofounder Simon Last, valued Imbue at more than $1 billion, making the company one of the only female-led AI research unicorns.
Competitors from OpenAI and Anthropic to Google are forging a different path, as GPT-4, Qiu and Albrecht vie to build massive AI foundation models. Imbue’s focus is an AI “agent”: a type of computing system that can emulate human decision-making to complete complex tasks. Chatbots like ChatGPT receive user queries and generate near-instant responses. Imbue’s agents will act like virtual research assistants who can crunch analyses, recommend follow-on experiments, and even set them up without supervision.
According to Albrecht, this kind of autonomy can be useful in a variety of situations, from biology research to travel planning and complex coding projects. Agents “go by themselves and do stuff,” Qiu explained.
“We believe AI has the potential to lower the barrier between ideation and execution.”
To build such an agent, Imbue gained access to 10,000 of Nvidia’s H100 GPUs — the same number of processors OpenAI used to train GPT-3. He released an open-source training environment for teaching such tools, called Avalon, last fall; More prototypes and releases are coming, its founders said, in the coming months.
Imbue is also strong on its journey to join the billion-dollar-startup ranks. The startup employs only 20 people. No demo of their agent is ready for public use yet, its founders said. And its lead backer, McCaleb’s Astera, is an unusual resource compared to the name-brand venture capital firms and big tech cloud providers that have fueled other recent AI projects — in part, Qiu and Albrecht claim, because nonprofits may be more likely to Patient with commercialization timeline. (says McCaleb Forbes (that he is still looking for “entrepreneur-style” financial returns in the long run.)
In a frothy AI market, a startup making hundreds of crores with no revenue is not unheard of. But it’s a bet to raise the stakes that draws more scrutiny on what Imbue does next. For Qiu, that goal is a lofty one: Do for AI agents what Xerox PARC Labs did for personal computers half a century ago.
At the time, computers were expensive and byzantine to use, but PARC researchers developed tools that made them accessible to non-technical people, and those tools eventually put PCs in our homes. Imbue aims to do the same for generative AI. “We believe AI has the potential to lower the barrier between ideation and execution,” Qiu said. “A truly personal computer that does things for you and frees people.”
iThe founders of mbue first met in 2014 at a conference at UC Berkeley. Qiu worked at Dropbox, where she started in business operations before becoming CEO Drew Houston’s first chief of staff. Before joining the data team at wealth manager Addeper, Albrecht was a cofounder and chief technology officer at several startups. Both had studied machine learning in college, with Albrecht co-publishing papers at the University of Pittsburgh, Qiu working on probability theory while MIT helped pay for her while creating and running high-frequency trading algorithms.
They tackled high-brow topics such as the future of human decision-making in an AI-dominated world. Soon after, Albrecht convinced Qiu to work on Ember Hardware, making his startup project virtual reality hardware, but the company never took off. After founding The Archive House, which became home to many early OpenAI researchers, the pair decided to use machine learning to solve a problem they experienced at the now-defunct Ember: inspiring talented people who aren’t looking for a job yet apply for one. Sourceres says his startup went through startup accelerator Y Combinator in 2017; Houston and his Dropbox co-founder Arash Ferdowsi invested as Sourceres raised a $3.5 million seed round after completing the program.
“Every decade or so a new era of computing begins, and everyone messes around together.”
The company did good business for a while, selling in the millions, before growth slowed. When several attempts to make the jump failed, founder and Threshold investor Josh Stein had a “mature conversation” about pivoting to another, potentially bigger idea. “There was nothing wrong with Sourcerace, but it was clear that breakout was not going to be successful,” Stein said.
In late 2020, Qiu and Albrecht closed Sourceress, returning about $4 million to investors and inviting those interested to roll over some of their investment into an equity stake in the duo’s new startup, Generali Intelligent. On paper, Imbue’s latest valuation makes those parts look profitable: “It looks like it’s going to be a very successful investment for us,” Stein said.
iIt’s a balmy August evening in San Francisco as Houston takes the mic to talk about all things AI with Qiu and Ali Rohde, who manage a small venture fund Imbue’s founders set up to back other early-stage AI startups. Dozens of attendees, mostly entrepreneurs and engineers, flock to the minimalist Mission District offices of productivity software company Notion to hear Houston speak.
This gathering, one of a weekly series, is reminiscent of The Archive’s old dinners, but on a grander scale. “Every decade or so a new era of computing begins, and everyone is messing around,” Houston later said. “Being in that community, it’s exciting.”
Houston didn’t expect much from his former employee recruiting automation into an AI research lab; Now, given the recent excitement about generative AI, he said the move has gone from looking like a “science project” to “right in the fairway”. He’s joined by a who’s who of backers in that fund, Outset Capital, which includes Quora CEO and OpenAI board member Adam D’Angelo, Hollywood mogul Jeffrey Katzenberg, and Google DeepMind head of policy (and former head of Dropbox PR) Dorothy Chou.
“We can either continue the research, while all these other labs are making great progress, or we can bet here and go for it.”
Despite their deep Silicon Valley connections, several prolific investors in the range — who asked to remain anonymous to speak freely — said they doubted the team’s credentials to run a serious AI research lab. Others disagree. One, who knows Imbue’s founders, discounted concerns such as VC bias to collect “baseball cards” or simply outperforming founders from many of the same prestigious backgrounds. Still, he wondered whether Imbue’s ambition to release the agent as a commercial project would stand out from other well-resourced labs in the long run.
Qiu and Albrecht are unfazed by any concerns about their team’s research chops. He points to a handful of employees with academic backgrounds in AI research as well as neuroscience and plasma physics, their breadth as a strength.
The lack of traditional investors on Imbue’s cap table (Stein’s threshold is the only one) has also raised questions. The founders say they deliberately did not hold any formal meetings with venture capital firms for Imbue’s recent fundraising, as they acknowledge that developing a proven commercial project could take years of their work. “I think it’s worth it to be a little outsider, to have a new look,” agreed McCaleb, his investor.
As for Imbue’s eye-opening assessment for such a new project, McCaleb said he was satisfied with the scope of the opportunity after seeing the demos that Qiu and Albrecht demonstrated. Maybe Finally create an agent; He went against “99% of the (research) efforts” he sees in other AI labs. (Qiu and Albrecht said they could not share the demo publicly yet).
“To actually take this research to the next level and see if we can build it and then produce it, you need a lot of money, because you need GPUs, right?” McLeb said. “We can either continue the research, With all these labs making a lot of progress, or we could bet here and go for it. So, it felt like the time to do it.”
Imbue may have its own public “aha” moment, similar to OpenAI’s popular launch of ChatGPT. These things take time. Qiu and Albrecht point to Cruz, the self-driving car business whose co-founders are now investors, and interest in the demo, which is now gaining more adoption a decade later. A mainstream agent like Imbue won’t take a decade, Albrecht claimed; But it won’t take months.
“We will actively explore,” Albrecht said. “We want to wait until it’s ready and think, this is really good, we believe in it, and it’s strong, safe and great before we put it out there.”