Database Reference
In-Depth Information
Jonas is an exemplar of the data scientist who is building the tools of tomorrow.
This comes through as he talks about creating instrumentation to record signals
from brain cells, his frustration with the slow pace of tool development in the past
decade, and his vision of having the right tools to analyze the deluge of brain data
that will be available in the future. His desire to build tools that help scientists and
non-scientists make sense of their world through data—together with his views on
how-quantitative-and-computational-the-life-sciences-will-become-going-forward—
energize his interview.
Sebastian Gutierrez:
You recently co-founded a startup and then sold it to
Salesforce. Tell me about this journey.
Eric Jonas:
Some friends from graduate school and I started Prior Knowledge
[P(K)] in August 2011 with the goal of building developer-accessible machine
learning technology. Our vision was building ubiquitous machine learning. We
were really inspired by companies like Heroku and Twilio and the way they
had democratized access to a lot of what at the time was fairly cutting-edge
technology. We felt it was crucial to preserve uncertainty—the ability to say
“I don't know the answer”—when putting this technology in the hands of
normal people. At the time, that was a really radical thought.
Certainly, the Bayesian statistical community had been doing this for a long
time, but a lot of machine learning methods that were out there were just
all about giving people an answer. It is important to ask whether you can
trust this answer or not, especially since often whether or not you can trust
an answer varies greatly on the question you happen to be asking. Machine
learning systems will quite reliably tell you that there are two genders, male
and female, but they won't always be accurate in predicting which one you're
talking to. So that was really the goal.
We did the normal startup thing—we built the product, we demoed and made
it to the finals at TechCrunch Disrupt, and we raised funding from venture
capital firms. Peter Thiel's group, Founders Fund, backed us. We then were
acquired in December of 2012 by Salesforce, where we found a great team
of people who very closely agreed with the vision of a ubiquitous predictive
platform, where at the end of the day saying “predict” should be as easy as
saying “select.” I was at Salesforce for over a year, where my team and I
continued to work on machine learning technology. I was promoted to Chief
Predictive Scientist of Salesforce and oversaw this area.
Gutierrez:
You have now left Salesforce to do computational neuroscience
research. Why neuroscience research?
Jonas:
The brain is fascinating because it computes. The liver is interesting
too—it's this complex metabolic soup, and when it breaks you die, so there
are people studying it. However, as a CS person, it's amazing that there's this
blob of goo in my head that somehow is doing all this intractable computing.
