The Allen Institute for AI (AI2) started its incubator two years ago, helping launch companies like Xnor.ai, Blue Canoe and WellSaidLabs. Their success has attracted funding from not just local Seattle VC outfit Madrona, but Sequoia, Kleiner Perkins and Two Sigma Ventures as well, resulting in a new $10 million fund that should help keep the lights on.
The AI2 Incubator, led by Jacob Colker since its inception in 2017, has focused on launching a handful of companies every year that in some way leverage a serious AI advantage. Blue Canoe, for instance, does natural language processing with a focus on accent modification; Xnor.ai is working on ultra-low-power implementations of machine learning algorithms, and was just acquired yesterday by Apple for a reported $200 million.
“We think the next generation of so-called AI-first companies are going to have to graduate into building long-term, successful businesses that start with an AI edge,” said the program’s new managing director, Bryan Hale. “And the people who can help do this are the ones who have helped build iconic companies.”
Hence the involvement of household names (in the startup community, anyhow) Sequoia and Kleiner Perkins, and Two Sigma Ventures from New York. Seattle-based Madrona also recently invested in AI2 company Lexion. It’s a pretty solid crowd to be running with, and as Colker pointed out, “they don’t often come together.”
“But also, they looked up into the northwest and said, what’s going on up there?” added Hale. Indeed, Seattle has over the last few years blossomed into a haven for AI research, with many major tech companies establishing or expanding satellite offices here at least partly concerned with the topic: Apple, Google, Nvidia and Facebook among others, and, of course, local standbys Amazon, Microsoft and Adobe.
Practically speaking, the new fund will let the incubator continue on its current path, but with a bit more runway and potentially bigger investments in the startups it works with.
“We just have a lot more resources now to help our companies succeed,” said Colker. “Previously we were able to write up to about a $250,000 check, but now we can write up to maybe $800,000 per company. That means they have a lot more time to build out their team, aggregate training data, test their models, all these points that are important for a team to raise a bigger, better VC funding round.”
AI2 prides itself on its large staff of PhDs and open research strategy, publishing pretty much everything publicly in order to spur the field onwards. Access to these big brains, many of which have bred successful startups of their own, is no less a draw than the possibility of more general business mentorship and funding.
Colker said the incubator will continue to produce three to five startups per year, each one taking “about 12-18 months, from whiteboard to venture funding.” AI, he pointed out, often needs more time than a consumer app or even enterprise play, since it’s as much research as it is development. But so far the model seems to work quite well.
“There are very few places in the world where an entrepreneur can come to take advantage of the brain power of a hundred PhDs and support staff. We’ve got a new research center with 70 desks, we’ve got plenty of space for those teams to grow,” he said. “We’re incredibly well-positioned to support the next wave of AI companies.”