Artificial Stupidity: getting state of the art results with less than ideal data sets, classifiers, or timescale

Dr Janet Bastiman

Reading papers on state of the art approaches using toy data sets is all well and good, but when you have a customer who needs nothing short of perfection, in very short timescales, without training data, what do you do? Say it’s impossible without lots of research or do you roll up your sleeves and McGyver a solution?

Companies don’t have time to nurture your research, they need results and they need them yesterday. How can you get a system that works reliably in less than ideal conditions while preserving your sanity?

In this session I’ll go through some of the practical things you can do to get the results you need in less than optimum conditions.

Objective of the talk

You’ll get a review of some of the situations in which my team and I have had to get results, broken down with solutions. As a practitioner you’ll take away ideas for how to get the results you need, as a technical manager you’ll get a feel for some alternate methods you might want to suggest.

Required audience experience

Implementation of AI solutions or technical managerial experience

Location: Date: October 1, 2019 Time: 1:35 pm - 2:20 pm Dr Janet Bastiman Dr. Janet Bastiman, StoryStream