March 30, 2017
I’ve been going to artificial intelligence conferences for a long time, and there are more and more every year. Most are focused in some way on the algorithms that power artificial intelligence. Train AI focuses on everything except the algorithm – the training data and feature selection and deployment issues that end up being 90% of the work. This is the conference by and for people that actually want to use artificial intelligence to tackle problems in the real world.
I started CrowdFlower because I needed more training data to make my machine learning models work. One of our keynote speakers, Peter Norvig had an especially big influence on me as one of the very first people to point out the outsized effect of training data size. I’ve always admired his paper, The Unreasonable Effectiveness of Data, where he explains the surprising results that using lots of training data can make simple algorithms look really smart. That was back in 2009 and machine learning is very different today. I’m really looking forward to hearing how his views have evolved.
I’m also excited to hear from Richard Socher. I first met Richard as the founder of MetaMind, one of the coolest startups in deep learning. Now he is Salesforce’s Chief Scientist and teaches the most popular class at Stanford: Deep Learning for Natural Language Processing. He’s both a very clear thinker and a very good teacher so I learn something new every time I talk to him.
Another speaker I’m especially looking forward to is Anthony Goldbloom. As the founder of Kaggle, a contest and community site for data scientists, he has direct access to the forefront of machine learning. No one has better data on some of the most interesting questions about what’s happening in the field: In which use cases is deep learning actually the most effective model? Or, what really matters in making a model work well? Anthony also has a great TED talk on which jobs we’ll lose to machines and which ones we’ll keep. It’s a topic that comes up a lot, but few have thought about as deeply or as technically as Anthony. If we’re lucky, hopefully we’ll hear some of his thoughts on this subject as well.
When I look at the overall speaker lineup, what most stands out is the diversity of industries that our speakers represent. CrowdFlower serves customers across so many different applications: from Dan Ruderman, a cancer researcher to Scott Prevost a VP of Engineering at Adobe to Barney Pell the founder of Moon Express to Shivon Zilis and James Cham, partners at Bloomberg’s venture arm. Surprisingly, we see a lot of the same challenges over and over across these different uses cases, so there should be a lot to learn from other industries.
In the past our conference sponsors have been mostly startups, and I’ve been particularly excited this year to see IBM Watson, Microsoft and Salesforce Einstein all sponsor Train AI in a meaningful way. These companies are all investing heavily in machine learning but have very different takes on the market and Train AI gives us a chance to talk directly to the people behind the scenes.
I’m looking forward to seeing you there.