Learning in the deep-end: my first tech conference

The O’Reilly Artificial Intelligence Conference in San Francisco was themed “Put AI to Work,” and designed to deliver practical, technical advice and training for organizations in this early stage of AI implementation. Most of the sessions explored applications of various machine learning methods within industries like healthcare, finance, and manufacturing where AI has started to take hold and the remaining algorithmic and computational challenges. However, there were a number of attendees looking for the broader view of what AI could do for their business, what kinds of questions they should try to answer with applied machine learning, and how to collect, organize, and deploy their data so that applying AI would create value.

Given the exponential growth of computing power and machine learning capabilities, the four AI conferences scheduled next year should expect an increasing audience of non-technical executives looking for ways to bridge their knowledge gaps around AI implementation.

Some highlights and themes:

1. Andrew Ng, former chief scientist at Baidu and founder of Google’s Brain Team, was the rock star keynote speaker of the two main days of the conference. His presentation expanded on his AI evangelism topic “AI is the new electricity” to touch on the features of the AI enabled product and the changing basis of competition for businesses. He drew and interesting parallel between what made an Internet company over the past 25 years and what will come to define an AI company in the near future. An unofficial recording of the presentation can be found here (see the last 6 minutes for the above).

2. Jana Eggers of Nara Logics lead an intensive seminar ahead of the main conference as well as a session during the main portion. Jana’s uncommon experience with a number of large-enterprise AI deployments has put her on the leading edge of making the business case for AI. She exposed in-depth Andrew’s point that X + AI does not equal a valuable solution, and broke down into key steps where (and with what kind of data) the opportunities for AI lie and how to overcome the major organizational barriers to developing them. One of the highlights of her session “It’s the Organization, stupid” was her emphasis that any organization looking to implement AI has to be ready to learn (from their data, about their problems, from the AI output) and keep learning (keep measuring, testing, and fixing). The AI-enabled company is a company with learning built into their core.

3. One emerging debate is how much AI solutions-makers should build transparency and verifiability into their designs vs. how much should executives shift their mindset to trust the AI. There are many conversations yet to come in this area.

4. The theme running through all of the presentations was data. On the other side of the big-data hype wave, everyone seems to understand that connectivity is not productivity, but many organizations still feel burdened by their own data. The good news is that rich, variable, real-world, yet small, data sets are becoming highly valuable as machine learning models and training methods are advancing. According to several of the presenters, future investors will assume a company is using the best available AI, it’s the value of the data its being applied to and the solutions or insights generated that they will spend money on.