Database Reference
In-Depth Information
a direct salesperson or a channel salesperson. While a different customer
would probably just care to know that this is a technical marketing salesper-
son, which is a completely different granularity. How we approached solving
this problem is that we ended up building an ontology and finding algorithms
that work with this ontology. This way we can find out the value of the net-
work granularity layer for each one of our customers. It seemed like such a
simple problem from the beginning, and the more we ran, it just became more
and more riddles to make it all work together like a well-oiled machine.
The other tricky thing about this ontology and getting value from the net-
work granularity layer is that we are rebuilding the models automatically on
an almost weekly basis. There are a lot of issues with that as well. As time
moves forward, the models change the distribution of how someone is scored
as people act and are seen acting. Even if the information about a specific
prospect didn't change, an A could turn into a B rank because we trained the
model differently based on global data changes. And because we deal with
small data, it can happen even on a regular basis, which means salespeople will
look at the new results and ask us, “What did you change?” So we have to take
into account the fact that we cannot change the distribution in a fast manner
because it has to make sense to the business side, which brings us back again
to the emotional AI problem.
Gutierrez:
How did you get involved with computer science?
Radinsky:
I came to Israel from Ukraine when I was 4. My mom bought me a
computer when I was around 5 or 6. She wanted me to learn and practice the
Russian language through the use of language learning programs on the com-
puter. She also bought me a lot of computer games. Some of these computer
games had math-related exercises. At one point, I had a trouble getting past a
really, really difficult level. I really wanted to move on to the next level, so my
aunt showed me how to write a “for loop” to solve the problem. In this way,
I learned to solve some of the exercises in an exhaustive kind of way instead
of doing the manual calculator way all of the time. That was the first time I
thought how amazing the ability to code is, “That's awesome. I can move to
the next level of the game faster by just using the computer.” This was what
started my fascination with computer science.
In addition to computers, as a kid I was also really excited about bioinformat-
ics. I was really interested in the fact that you could take all genetic data, actu-
ally it it into computers, and solve many problems that look unsolvable before
that, and reach medical discoveries. You could potentially build a combination
of a human and a computer together. I took part in the Technion External
Studies program when I was 15 years old, which allowed me to start taking
college level classes while still in high school. And once I started studying at
the Technion University, this was what I wanted to do—study bioinformatics.
Before my studies started, at the age of 14, I went to a research camp. At the
camp, each one of us selected research he or she wanted to lead—I chose
