http://techcrunch.com/2012/08/21/vicarious-good-ventures-funding/

Vicarious, a startup that says it’s “building software that thinks and learns like a human,” has just raised a $15 million Series A.

The round was led by Good Ventures, the investment firm founded by Dustin Moskovitz, who also co-founded Facebook and Asana. (The firm’s profits will be donated to the Good Ventures Foundation.) Founders Fund, Open Field Capital, Steve Brown, and Zarco Investment Group participated too.

Vicarious launched in February 2011 with funding from Founders Fund, Moskovitz, Adam D’Angelo (former Facebook CTO and co-founder of Quora), Felicis Ventures, and Palantir co-founder Joe Lonsdale. Since then, co-founder D. Scott Phoenix tells me that the company has been in research mode. The research has resulted in a system that’s supposed to interpret the content of photos and videos in a way that’s similar to humans, and which is powered by the company’s “key innovation”, the Recursive Cortical Network.

Ultimately, Phoenix says the technology could be used in “almost every industry,” including robotics, medical image analysis, and image and video search,. But that’s a ways off — Phoenix and his co-founder Dileep George say they’re still deep in research and development, and that the funding will be used to expand those R&D efforts. Developing products that commercialize the technology is still several years off, George says.

“Based on our experiments in the last year, we are very optimistic about our rate of progress,” he says. “At the same time, this is a very challenging problem. We are not getting too excited about how productize things. We’re testing everything very carefully.”

You don’t see too many venture-backed software companies spending years on research nowadays, and Phoenix says he was lucky to find investors who share his big vision — to use AI to “help humanity thrive.” The investors at Good Ventures and Founders Fund have a “natural affinity” for that kind of talk (Founders Fund’s Peter Thiel, for example, has been pretty vocal about what he sees as a lack of transformative innovation), but Phoenix says it’s “very different from the language that a lot of other investors speak.”

Discussion on Hacker News

Meta: this is my first post here, let me know if I am doing something wrong or if I shouldn't post this here.

 

New Comment
13 comments, sorted by Click to highlight new comments since:

Welcome! And I appreciate that you included a meaty blockquote in your post instead of relying on the title alone to make me click out.

Would be nice to have details of their algorithmic approach, instead of some nebulous buzzword like 'Recursive Cortical Network'. I suppose it does hint somewhat at neural networks...

Their website also seems to emphasize the wrong thing - emphasizing the potential of visual processing algorithms and such. I would be more worried about whether their team is smart/visionary/revolutionary enough to make significant headway on such a difficult problem. Because they're emphasizing the 'wrong' things, it sets off my 'Solyndra' alarms.

Agreed. Just giving their algorithm a fancy name doesn't mean that they've come up with something signficantly better than the ideas in the countless papers in Computer Vision... the only evidence for that seems to be the 15M they just raised (but for that it seems to be enough if you have a good implementation with near state-of-art performance and a workable business idea. Or is it?)

It's fine to skip it if you want, but new users are encouraged to check out the welcome thread.

"thinks and learns like a human"... do they mean "using some kind of neurons"?

Otherwise, having had a look at their possible applications, I think they might have some cool narrow AI vision system, but aren't even aiming for AGI (and they didn't even claim that, except for the above line). 15M for any AI company still sounds good, but I don't even have a guess how much is for example Google spending on stuff like that?

How does Thiel square this with his support of SI?

I assumed Thiel had enough money to throw a million here or a million there at anything he thought was interesting, and that that was what he did.

Brainstorm time!

He doesn't. Always an option.

He thinks it will have low risk of self-improving AI, and will have positive impacts directly or from innovation or from changing culture.

He thinks it will lead to earlier self-improving AI, but thinks SI will be ready.

He thinks it will lead to earlier self-improving AI, and thinks of SI as just a hedge.

Could influence, but to me he have strong sense with projects who are apparently radical, for scientific reasons or for academic ones.

I thought Thiel donates to SI simply because Yudkowsky is his friend.

Maybe he trusts himself to see when things are getting risky, and having a share buys him leverage over the direction of the company.

Alternatively, he simply thinks of SIAI's stuff as important background research to have in place, but doesn't think we're anywhere near building AGI anytime soon.

I haven't found any reference to "recursive cortical networks", other than in their press release.

Anyone know what they're talking about? Do they have a patent pending, or why the secrecy?

I don't see what's "strong AI" about it from the top-level description. Perhaps their machine learning algorithm will become widely used. There are a lot of machine learning tools already.