Database Reference
In-Depth Information
In the past, it was really surprising to me how much correlation you have from
different data points with events that are happening in the world and that all
the data is already there.
Gutierrez:
How did you develop your personal philosophy on working with
data?
Radinsky:
Basically, I've by working with a lot of very talented people. I think
the main person who affected my views was my advisor at Technion, Professor
Shaul Markovitch. He gave me most of his habits about how to approach
problems. For him, most of the things he was talking about were: Is it cre-
ative enough? Is it an interesting problem nobody's solved before? The way
he would approach that was, first of all, “Let's have data. Let's understand that
we know how to test the novel things. And after that, let's try a different
hypothesis.”
The next person who really affected me is Susan Dumais, a Distinguished
Scientist at Microsoft. What I really love about the way she thinks is that she
isn't about “Let's do something creative.” Instead, she's about “Let's make a
hypothesis that is interesting” or even a psychological hypothesis that's based
in data to see whether it works or not—to try to find out insights. My advisor
was more of an algorithm person: “Oh, let's make a cool algorithm.” And he
wasn't purely problem-driven. Susan Dumais was the first person I worked
with who was really problem-driven. She would say, “We develop insights from
the world around us. We are scientists. Even if we're working in computer sci-
ence, we're still scientists. Somebody who's working in physics has hypotheses
and tests the data. This is the same thing that we do. We have hypotheses and
we test the data.” So she really pushed me to be problem-driven.
And at the next level would be Eric Horvitz, who is more about making some-
thing of value that's going to change big things, especially in the medical com-
munity. Use those hypotheses to make something that is actionable. I'm really
inspired by Eric Horvitz. We've worked together a lot and I really love his
work in the medical domain. Especially, I was really impressed with how he
took all the creative thoughts that people have at Microsoft and used them
to try to predict the different side effects that drugs can have. This is great
research for drugs and the different side effects that they are having. For me, it
was pretty amazing how you can use all these things to actually help medicine
today. More than that, I'm pretty inspired by the fact he has done a lot of work
about decision making, which I like, especially as I'm still interested in how
decision makers make decisions.
Gutierrez:
What does the future of data science look like?
Radinsky:
I think that a lot of things in data science have been commoditized
into very simple tools where you don't have to understand them in depth to
use them. This is how I think about how our slice of the world has developed.
At the beginning, we had hardware and a lot of people worked in hardware.
