Photo: Yolanda Sun
Last week, I spent a couple of days in San Francisco for the IBM Watson Developers Conference and our nation’s largest AI business conference: AI World Conference and Expo. I met a handful of the very brightest - and most prominent - members of the AI community: Qualcomm Ventures Partner, Patrick Eggen; NVIDIA VP & GM, Jim McHugh; PwC Global & US Consulting Analytics Leader, Paul Blasé; IBM Fellow and IBM Watson VP and CTO, Rob High; Baidu Senior Researcher and Head of Systems, Gregory Diamos; Artificial Solutions CEO Lawrence Flynn and CMO Andy Peart; Steve Ardire, Advisor to Software Startups to name a few. They blew my mind with what they had to share!
I consolidated key highlights from my chats - and general sessions - into 10 things you need to know about the enterprise AI world. Enjoy.
1. Investing: The ideal AI B2B Enterprise seed investment for a VC involves startups who’ve moved beyond a science project, and whose products and services offer immediate commercialization opportunities and easily translate into other markets.
2. Compute Power: AI deep learning requires an increased amount of computing power. To address the need, NVIDIA built the DGX-1 supercomputer which packs the horsepower of 250 servers.
“We now have a new computing model to power AI and solve what was once unsolvable, with superhuman speed and intelligence.”
- Jim McHugh, NVIDIA VP and General Manager
3. Manpower: AI has historically required PhDs and experts to understand and use it. Today, the tech industry is developing AI platforms and tools (e.g. Microsoft’s and IBM’s open source AI toolkits and APIs) that can be operated by non-specialists.
“It’s like the Ratatouille movie, but instead of ‘anyone can cook’
it’s ‘anyone can do machine learning.’”
- Amin Mintrach, Yahoo Research Scientist
4. Hype: AI is not magic and AI is not the right approach to all enterprise challenges or goals.
“Focus on your company's challenges and what it is that you want to achieve.”
- Beena Ammanath, General Electric VP Data and Analytics
5. Quirks: Here are three to name a few: 1) It’s easy to get started with AI but it’s hard to get it right. 2) It’s a challenge to get both accuracy and an explanation. 3) The AI black box effect is a major problem - this is when you add inputs and get outputs without really understanding what happens in-between.
6. Human Potential: AI’s full potential is not to simply automate repetitive tasks to achieve higher levels of efficiency, but to, more importantly, augment our ability to elevate our thinking (e.g. see things differently, notice vulnerabilities not noted before).
“The biggest constraint is people’s ability to understand
the potential of what they can do.”
- Rob High, IBM Watson VP and CTO and IBM Fellow
7. Human-like AI: Coming soon. Expect AI to sound and converse like a human. Intu will help with this. Think physical, cognitive, and emotional. It’s happening.
"The future looks like the movie, “Her,” except that at the end she doesn’t leave us.”
- Dave Parsin, Artificial Intelligence VP of North America
8. Job replacement: It’s going to happen across the board. You can see how replaceable you are - thanks to Mckinsey - here.
9. Success: The AI requirements for success include, but are not limited to, 1) data (a lot of “good” data), 2) a problem to solve (the problem must have a monetary tie to the enterprise - e.g. growth, risk reduction), and 3) maintenance and monitoring of AI systems.
“Think of AI like a rocketship, where data is the fuel and
neural networks are the engine.”
- Greg Diamo, Baidu, Inc Senior Researcher
10. Enterprise AI Insight: The Bloomberg Beta team did a great job consolidating the enterprise AI ecosystem here and summarizing their observations here!
"Startups, consulting orgs, big players all want and try to do this but it's complex like playing 3D chess, so we have to be an extreme pattern matcher to plan, anticipate, execute many moves ahead to win."
- Steve Ardire, Advisor for Software Startups
With so many elements to AI, I agree with Steve Ardire, Advisor for Software Startups, and Paul Blase, PwC Global and US Data and Analytics Consulting Leader in that it’s helpful for enterprises to engage with partners who can provide: AI consulting and quant resources; bring AI apps and data to the table; provide education; and serve as a lab partner along the way. Pick wisely and welcome to the world of AI.