To shed light on this major transition in the enterprise, I spoke with Sam Charrington, founder of TWIML. Charrington hosts This Week in Machine Learning & AI, an influential podcast that features top industry experts.
Charrington has organized a major AI conference: TWIMLcon: AI Platforms, October 1-2, 2019, in San Francisco. (Datamation readers get a twenty percent discount by entering DATAMATION20 when registering.)
The TWIMLcon conference will feature a long list of industry leaders, including Uber’s Franziska Bell, Twitter’s Abhishek Tayal, Cloudera’s Amr Awadallah and Landing AI’s Andrew Ng.
The event will include sessions by experts on everything from scaling ML to the ethics of AI. It will likely be a networking extravaganza for deep learning geeks.
See below for a transcript of highlights of my discussion with Sam Charrington.
Artificial Intelligence Trends: Sam Charrington on AL and ML
Investment in AL and ML
“Most enterprises, large enterprises that are technology forward, have been working with machine learning and AI for several years now. I was just on a call with someone at one of the leading consulting companies, who works with a lot of enterprises. He was talking about one of his customers in oil and gas. They’ve been doing these machine learning proof of concept projects for the past three, four years, to the tune of 250 individual projects and $150 million in investment.”
“There is a ton of model shelfware out there because a lot of the initial impetus on the part of these organizations that spin up machine learning projects is to learn and understand so that when the transition happens, they’re not so far behind that they can’t catch up. So there’s a lot of experimentation ‘science projects,’ we call them sometimes, that are happening.”
ML Has Demonstrated Business Value
“About a year ago, there was this silent echo across enterprises: ‘Okay, we get it. We see that there’s value here. You’ve demonstrated that we can use machine learning to provide business value. How do we actually go that last mile, get these models into production, and furthermore, build out processes, infrastructure, and platforms that allow us to scale this across the enterprise?’
“Because most data science teams don’t have the resources, the people resources, to take on all of the opportunities that they’re presented with. They’ve got to be very selective and manage a portfolio of opportunities, but there’s so much more impact they could have if they could build an engine for getting models out and into production and providing value.”
AI and DevOps
“All of the DevOps topics have analogues on the machine learning side – MLOps is what you hear. Often, we talk a lot about machine learning platforms, continuous delivery of models, continuous monitoring models, once they’re in production; these are all very relevant topics that many organizations are trying to figure out currently how to do.”
The Challenge with Hiring Top Talent
“It’s still very difficult for organizations to compete with the Googles and Facebooks of the world for top-tier machine learning, AI, and data science talent.”
Deep Learning Driving AI
“Artificial intelligence has been around for a very long time. As an industry, we’ve gone through several ‘AI winters,’ with the crest and the waning of interest in AI and funding available for AI projects. But the most recent enthusiasm and excitement about AI has largely been driven by what’s become possible with deep learning.”