Vicarious announces $15 million funding for AI software based on the brain
August 24, 2012
An abstract visualization of the neuromorphic Recursive Cortical Network (credit: Vicarious)
Vicarious FPC Inc, an artificial intelligence company that uses the computational principles of the brain to build software that can think and learn like a human, has announced a $15M Series A round of financing for development of machine learning software based on the computational principles of the human brain.
The research at Vicarious is expected to have broad implications for robotics, medical image analysis, image and video search, and many other fields.
“Building machine intelligence is one of the most important and challenging problems humanity has ever faced. Advancements in neuroscience, probabilistic models, and computing power are enabling new strategies for AI research,” said Vicarious co-founder Dr. Dileep George.
A new computational paradigm: the Recursive Cortical Network
Since its launch in February 2011, the company has developed a visual perception system that interprets the contents of photographs and videos in a manner similar to humans. Powering this technology is Vicarious’ key innovation: a new computational paradigm called the Recursive Cortical Network (RCN).
RCN is an extension of the motivation behind hierarchical temporal memory (HTM), a machine learning model developed by Jeff Hawkins and Dileep George of Numenta, Inc. — to model some of the structural and algorithmic properties of the neocortex.
“HTM was an important effort,” George told KurzweilAI. in an email interview. ”Much like Poggio’s foundational HMAX model, HTM had the right high level goals. But, when you dig into the algorithm level, you’ll see that HTM implementations hadn’t solved the problems of information representation in the hierarchy.
“This led to inefficient learning and scaling issues. The mathematical formulation of HTMs were made up of these “blocky” nodes with boundaries that restricted information transfer between adjacent nodes, which is not very effective in dealing with domains like vision and sound.
“After I started Vicarious, I had the freedom of a clean slate and could look at the problem with fresh eyes. My goals have always been to embody the computational principles of the brain in a mathematical model, but RCN is a ground-up rethinking of what kind of algorithmic approach is necessary to solve the problem.”
Investment by leading venture funds
The investment was led by Good Ventures LLC, a for-profit investment firm founded by Facebook and Asana co-founder Dustin Moskovitz that will donate all of its earnings to the Good Ventures Foundation. Also participating in the round are veteran institutional investors Founders Fund and Open Field Capital, as well as Vicarious angel investors Steve Brown and Zarco Investment Group.
“Vicarious is bringing us closer to a future where computers perceive, imagine, and reason just like humans. We are proud to support Vicarious in its quest,” said Peter Thiel, partner at Founders Fund.
“We are honored to pursue this long term goal with the support of investors who share our deepest values,” added Vicarious co-founder D. Scott Phoenix.
Vicarious was founded in 2010 by D. Scott Phoenix and Dr. Dileep George. Before co-founding Vicarious, Mr. Phoenix was Entrepreneur in Residence at Founders Fund and CEO of Frogmetrics, a touchscreen analytics company he co-founded through the Y Combinator incubator program. Previously Dr. George was Chief Technology Officer at Numeta, a company he co-founded with Jeff Hawkins and Donna Dubinsky while completing his PhD at Stanford University.
Comments (20)
by mlohbihler
I took the time to watch the presentation, and was saddened to hear Dileep still talking about image recognition as a starting point, like gliding was a starting point for flight. This is the wrong approach, as i already explained over three years ago: http://gameofid.com/blog/the-false-lure-of-image-recognition.
by mlohbihler
AGI *is* way off in the future, speaking as someone who has been “in the industry”, as it were, for over a decade. Each new idea appears to be a move in the right direction, but i’ve watched as these ideas repeatedly fade out during the scaling attempt. (See my recent recount of such here: http://gameofid.com/blog/the-full-htm). I absolutely do hope that Dileep is on to something, but until he shows us something real, i’ll continue my own tinkering. And even if he does produce something useful, it will only incrementally substitute human intellect, likely over a period of decades. The brain, after all, is not a single algorithm.
by Uh-oh
Oh well, they are on the right track with networks but still focus on “developing algorithms that mimic the function of the human brain”. Interesting at first, but then again, this is just flushing money down the drain for outdated technology. All you get with algorithms are useful tools, not free-thinking AIs.
by Cybernettr
I see they are backed by deep pockets like PayPal and Facebook. Yes, it looks like we are indeed out of the “AI Winter.”
by trakk
Until we fully understand the brain, we cannot develop AI that fully mimics it.
by John Middlemas
Scott Phoenix mentions a “common set of principles” for the neocortex in the video around 5:39, however this is negated according to http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1569491/ where it states “Although the column is an attractive concept, it has failed as a unifying principle for understanding cortical function. Unravelling the organization of the cerebral cortex will require a painstaking description of the circuits, projections and response properties peculiar to cells in each of its various areas.” Many have mistakenly believed, and still believe, a highly influential paper published in 1980 called “The basic uniformity in structure of the neocortex” which was mostly incorrect information, http://www.pnas.org/content/105/34/12099.full
by star0
That may as well be; but I gather that their system will bear as much resemblance to an actual brain as airplane wings do to the wings of a bird.
“But where are the synaptic clefts? And what about the ganglion cells and pyramidal this and axonal that? And we don’t yet even know… How can you claim to have… ” come the criticisms, which all miss the point. The point is to use current neuroscience to get some ideas for what sorts of algorithms the brain might be using to learn, then to try to build IDEALIZED mathematical models to do the same, and lastly to test and refine those until they get something useful. If it so happens that the brain uses a large variety of completely different and incompatible learning mechanisms, then I’m sure Vicarious will adjust their research program according to that new information.
I remember the first company that Dileep was a member of, called Numenta; and I remember seeing some of the demonstrations of just what their technology could do (things like tracking people in surveillance videos and recognizing symbols in noisy environments and under various transformations, as I recall). It was very impressive, but not perfect. If what he has produced now is even better than that, then he’s already got something quite good, in my opinion.
by hubris01
Sigh. There is zero mystery why airplanes and bird wings work; it’s called the Bernoulli’s principle. Although they do in fact adapt it in different ways, the fundamental principle is not an idealized model. It is a physical implementation based on a well established principle that was available to the Wright bros and other competing efforts at powered flight. Neither Numenta nor Vicarious (nor any Turing equivilant instantiation for that matter) implement anything remotely like the complex superimposed chemical and electromagnetic dynamics in actual brains.
Sure, they will make some tools, maybe even saleable tools. But it will not realize their goal of implementing animal, let alone human, sentients.
by star0
You’re reading too much into what I wrote about the bird wings versus airplane wings. I would also say that it’s not trivial to work out how airplanes produce lift based on an understanding of Bernoulli’s principle. See:
http://www.grc.nasa.gov/WWW/k-12/airplane/bernnew.html
and the following one about Einstein making a mistake himself:
http://wrightstories.com/einsteins-wing-flops/
(The article itself seems to be slightly erroneous, in postulating the incorrectness of applying Bernoulli to wing design.)
Yes, I agree completely about the “complex superimposed chemical and electromagnetic dynamics in actual brains”. And, yes, this is about making tools (at least for the next 5 years or so), not about making an AI or trying to reproduce all the complex processes that go on the brain, at least that I can see. Maybe 15 years from now the project might have morphed into one to build an AI; but that doesn’t look to be on the near-term horizon, according to them.
by star0
One more thing: I would argue that Hawkins’s HTM model plays the role of the Bernoulli Principle in this case. Now, there is a world of difference between them: Bernoulli’s Principle was well-established at the time the Wright Brothers developed their prototypes, and it had a precise mathematical formulation; HTM, on the other hand, is more an inspired guess (it’s better than that, in fact), and may ultimately be false, even when applied to the visual centers of the brain — but it still provides a direction in which to proceed to build a neural network to learn, for example, how to recognize objects.
by snake0
It looks like advancements in neuroscience are going to be the key to getting out of the AI bottleneck we’ve had since the 80′s.
by star0
I wrote in the forums that 15 million sounds like a lot for the Series A funding of a small software company (I base this on what I read on wikipedia about the typical size of series A for a pure software company). I’m guessing, then, that the “show and tell” to funders must have gone really well — makes you wonder just what they’ve got…
by Jamie_NYC
Or, it may be that they don’t expect any commercial product to emerge for a number of years, so need a substantial cash reserve to keep operating…
by star0
But is that typical of Series A funding? I thought Series A was for just a few years — wikipedia says 6 months to 2 years is typical.
by MrFriendly
Nice talk in the video. I just wish they could show more of their actual progress, but I don’t blame them for keeping a lid on most of it. It is best that they don’t try to commercialize too soon. Very smart.
I love to see smart people reaching for the stars like this. I’m also really glad that we’ve come out of the AI winter, and that serious money is being spent on these pursuits. Honestly, ANY amount of progress in computer vision can help change the world for the better, given what it could do for web search, manufacturing, robotics, etc.
I wish them much success.
by Bri
Are you sure that it will take till 2045 for AI to think like a human? Looks to me like it could be the early twenties, maybe even sooner!
by Editor
I don’t know who you are quoting.
by Jared
I thought Ray Kurzweil said that Ai would match human intelligence by 2029 and 2045 would be the approx time when the singularity would happen. I agree its way cool, but Ray predicting these elementals happening now and into the late 2020′s. It would be cool if it happened sooner tho, but I’ll stick with Ray’s predictions.
by Bri
Ray has his hands on the pulse, so to speak. He has access to information that you can’t get in press releases. Paul Allen feels Ray is overly optimistic. Mr Allen has dumped 500 million into brain research and also has a unique vantage point. I only have the impressions that I get from reading reports like this. To me it’s going so fast that it feels like it could be tomorrow that human AI will come from some obscure lab. I think the problems with AI are being affected by human brain complexity. Researchers working with octopuses are startled by thief intelligence.many live only one year. They report that some octopuses will break out of thief tanks and eat a fish in another tank, then sneak back into thier own. That’s some very impressive problem solving. Why bother sneaking back into thief tanks? Why not make a run for it, so to speak? That not a mammals brain, and I’d like to see a one year old human do that. Just makes you wonder if they can’t see the forest for the trees!
by Bri
Woops my bad! I was thinking how some people think it’s way off in the future.. Just seems to me to be around the corner. 2045 is the singularity. It just poped into my head. Too early in the morning!