Database Reference
In-Depth Information
to perform a research on how natural compounds affect the proliferation of
cancer cells, specifically prostate cancer cells. We studied a lot of cells and
tried to identify how they affect cancer. And working with this data, we had
a lot of images of how those cancer cells grow. This was the first time I did
image processing, because I didn't want to count the cells by hand, as there
were something like five thousand cells to count in each image. So what I did
was write a system to do this image processing for me, which gave me some
free time to go to the pool instead of doing tedious work.
While doing this work at camp, and then my studies in the Technion External
Studies program, I started reading more and more about artificial intelligence.
I thought to myself, “Wow, that's amazing! We have so much data around us
that we can actually leverage.” I finished high school and part of my bachelor's
degree, and then I went into the army. There I was a software engineer for
three years, mostly doing security-oriented work. Then I came back to the
university to finish my bachelor's degree.
Gutierrez:
When did you realize the power of data?
Radinsky:
In 2011 there was this event in Beebe, Arkansas, where something
like five thousand dead birds fell out of the sky. It was very interesting because
it was close to the dates of the end of the Mayan calendar, so everybody was
thinking that this was a sign of the end of the world. Not only that, just a few
days before, hundreds of thousands of fish washed up dead on the shore in
a nearby part of Arkansas. And all the newspapers were reporting that there
is no relation between the two incidents. Nobody could understand why the
birds died or the fish.
I wouldn't say that this was my first passion for data, but this was the first
time I actually understood that working with data was like an adventure. For
me it was, the “well, what's going on here?” moment that pushed me to do
additional investigation of that. So I started using Google Trends to look for
peaks of searches for bird death and fish death to see when they tend to
happen. I noticed that they tend to happen at approximately the same time.
The data I was using at that time was the images of Google Trends, because
they didn't have an API yet. I actually had to take the image and apply image
processing to them in order to scan all of the lines involved so that I could
have the data behind the charts. It was a lot of work to do something pretty
simple. After extracting all the data, I wrote a system that looks for correla-
tions of events that tend to happen before people search for birds and fish
death in an unusual way.
What the system automatically found out is that in the location where people
search for oil spills, even up to six months before, those two queries tend
to peak together. The thing with oil spills is that they cause oxygen deple-
tion, which is the number-one cause of fish deaths. I thought to myself, wow,
interesting—that might be relevant. I don't know if this was actually what was
