Database Reference
In-Depth Information
The data microscopes project that DARPA is funding is all about building
exactly these sorts of tools to let you see things in data that you couldn't see
before. A great deal of data analytics startups these days are like data visualiza-
tion technologies in that it's great when you can think of the questions to ask,
and then ask and visualize and plot the conditionals and these sorts of things.
However, for a great deal of data regimes, we're far beyond that.
When you start looking at these kinds of Grande Data problems, and it's too
much to visualize, and it's not enough data to do something Google-scale, and
you don't even have the resources to do Google-scale—well, what do you
do now? More linear regression? I really think that's going to end up being
the future, especially for the sciences. It's inevitable. And I think you're going
to eventually see these sorts of data microscope techniques being taught to
undergraduates. We're going to see this real transition.
Gutierrez: What data sets would you love to see developed?
Jonas: There are two data sets. One is that I would like to have all of the
connections between every cell in the brain. This would be the spatial
locations and connections of every cell in the brain. Two is that I want the
time series from every neuron in the hippocampus. We have to start building
these sorts of high-throughput neural data sets. I'm not necessarily content
with these being in animal models, as they're going to be for a while, but we'll
get there eventually because the systems are there. The data is there; it's just
currently inaccessible. We have to change that. Fortunately, there's more and
more interest. Somewhere in my brain, there's some glob of goo that knows
my phone number. We have no idea what that is, what that looks like, how
it even works, and that's ridiculous because it's in there. There's some circuit
in there. I want to understand that, and I want the data to exist to help me
understand that.
Gutierrez: Early in your PhD you built a device to measure this data. Have
you or others thought of pursuing this?
Jonas: It is something I've thought of pursuing. The question is partly one of
comparative advantage. I think that this space is large enough. What I really
want to get to is—if you can record all those neurons, how do you then go
use that technology to make $10 billion? Because then we can let the capital-
ist innovation machine do its work. However, what we're currently on right
now are rats. No one really wants to read the mind of a rat. Even pharma
companies have no interest in reading the mind of a rat because a rat is basi-
cally too big of an animal to do large-scale experiments with. They like mice
because they're small, and even then, mice studies are horribly expensive from
pharma's perspective. We'll get there eventually. Hopefully, the hype doesn't
kill it first. But, I can imagine going back on the instrumentation side. I've spent
a little bit amount of time over to Berkeley's AMPLab, working with some
people there that do compressed sensing, trying to feel out these areas. We'll
see. Hopefully there are lots of opportunities there in the future.
 
Search WWH ::




Custom Search