I’ve been an attendee and fan of the Structure conference series since the very first event in 2008, so I was excited to attend the rebooted Structure Data conference last week. The team over at Structure pulled together a great agenda, and I enjoyed the reincarnated data event much more than previous editions held in New York. Here are some thoughts and highlights from the event!
GE and the Internet of Things
One of the opening sessions of the conference was a conversation with Bill Ruh, CEO of GE Digital and the chief data officer for all of GE, and Steve Lohr of the NY Times. GE has been a poster child for embracing digital technology for quite some time, and they put a real stake in the ground last fall in forming GE Digital and announcing their intent to become a top 10 software company by 2020.
The company’s software portfolio includes Predix, a cloud-based platform for big data and analytics based on Pivotal Cloud Foundry and embraced by GE businesses and 3rd parties alike.
Bill mentioned an interesting energy use case that called for mini compute clouds to be deployed close to a wind farm to process their raw data and use it to manage the performance of the individual turbines. (Note: Predix wasn't specifically mentioned here, so this use case could be based on a different technology.) We'll see a lot of this edge processing pattern deployed to support #IoT applications.
GE using "mini-clouds local to the wind farm" to optimize performance of each wind turbine, incr efficiency > 20% #structuredata— Sam Charrington (@samcharrington) March 9, 2016
Bill made an interesting point about how utility economics were taking over the industrial world. It’s not just us, IT folks.
In other words, shifting from selling assets to selling outcomes.— Sam Charrington (@samcharrington) March 9, 2016
One of the things I got from the conversation with Bill is that we’re just starting to wrap our heads around what AI means. Steve asked what fundamentally new things has data allowed GE to do. The examples provided—the one that jumps out at me is using AI-powered robots to clean railroad tracks—while certainly being innovative were still in my view incremental. In that case we’re replacing a human worker with a machine doing the same task, albeit perhaps better, faster and cheaper.
My take on that answer: No. Robots fixing things vs humans is still optimization. #structuredata— Sam Charrington (@samcharrington) March 9, 2016
This is not to take away from the impact of what GE has been doing with data and analytics, but I’m confident that in a few years time we’ll look back and see companies like GE placing huge bets on AI enabled business models that simply weren’t previously imaginable.
AI: Context and Implications
AI was also the subject of discussion in great conversations with Andrew Ng of Baidu and Jeff Dean of Google.
Andrew is very bullish on self-driving cars, but thinks we’ll need to significantly manage their context in order for them to be useful in the short term. In other words, while he doesn’t see a fully autonomous vehicle being let loose on today’s roads within five years, he does believe we can make modest changes to our infrastructure to make them viable in that time frame. He didn’t really expand on what he thought those changes would look like, but you can imagine things like separate lanes for self-driving cars, parking lots specially designed for autonomous vehicles, etc. There’s an analogy here to electric vehicles: they didn’t really become a viable choice for many drivers until the auto makers along with governments stimulated the deployment of charging stations.
This notion of context and how important it was to effective use of ML and AI was a theme that recurred in the event's discussions.
Your kids spend 3 yrs of life in cars, and cars reduce life expectancy by 3 yrs. So self-driving cars gives them 6 yrs back. #structuredata— Sam Charrington (@samcharrington) March 9, 2016
Jeff spoke about Google DeepMind’s AlphaGo and it’s match against Korean Go master Lee Sedol as well as the massive growth of neural networks and deep learning technologies within Google.
Based on my summary so far you’d think AI was all that was discussed but that certainly wasn’t the case. There were a bunch of conversations talking about big data infrastructure as well. All three of the major Hadoop vendors—Cloudera, Hortonworks and MapR—were represented on stage.
A couple of conversations brought up use cases and the perspective from the vendor community seemed to be that the market was maturing such that enterprises were starting to look beyond the low-hanging fruit towards more interesting (read: transformative) use cases.
June Andrews from Pinterest and Josh Wills from Slack offered some nice insights on the relationship between algorithms and infrastructure in their discussion on “unicorn-scale” data science.
Netflix’s Eva Tse spoke about that company’s approach to big data infrastructure. Particularly interesting is their approach to big data security, specifically that they’ve imposed no barriers to access. This is especially interesting given the high level of interest and investment in big data security tools we’re seeing right now. I think Netflix has it right here, but this approach requires sophistication and won’t be right for all companies and use cases.
Personalization is a killer app for machine learning and AI, and offers both low-hanging fruit and opportunities for transformation. Representatives from IBM Watson, Yahoo Labs, Pandora all spoke at length about how (and why) they’re taking on personalization-related challenges with AI and machine learning technologies.
Resiliency is big issue for personalization. Notion of pers8ion getting messed up b/c kid used it resonated strongly here. #structuredata— Sam Charrington (@samcharrington) March 9, 2016
Challenges Adopting Advanced Data & Analytics
The business and technical challenges faced by those adopting advanced data technologies were raised throughout the event. Dan Wagner, CEO of consultancy Civis Analytics gave voice to concerns he hears frequently from the business leaders he works with. Interana’s Ann Johnson made a similar point. In both cases the issue is analytics technologies being sold as “magic,” leaving the business holding a bag that can’t be put to practical and effective use.
Over on Twitter, Jake Kaldenbaugh took this a step further (and kicked off an interesting convo in the process) by placing some of the blame here on enterprise startups that focus to much on technology and not enough on services and business solutions.
This is the hard part that many enterprise tech startups try to avoid, but maybe can't. https://t.co/ifVIx5o2G4— Jake Kaldenbaugh (@Jakewk) March 10, 2016
There were a lot more great discussions at the event than I could possibly cover here. If you liked this post and these tweets, please do follow me on Twitter.
I’ll leave you with this gem of a quote by McLaren Group CIO Geoff McGrath.