Strata Data Keynotes Day 2 on Dimming the Noise, Faster Data, and Becoming Boring

Here's my synopsis on a great start to day 2 of the Strata Data Conference. It began with a series of thought provoking topics; eliminating noise; driving faster data; successful data science by becoming boring.

Eliminating Noise and AI for Compliance

Amber Case of the MIT Media Lab talked about the noise, and the abundance of alerts that surround all of us. In our jobs... On planes... In our hotels.... High frequency content can boost salt and sweet flavors on airplanes (which is good) and on road trips (which gets you to eat more snacks). Her principles for sound design in products include: (i) Use better hardware that have better frequency ranges, (ii) Embed more information into sound   (iii) Remove mechanical noises (example: Blendtec blenders) (iv) a lower level sound conveying information like Star Wars' R2D2 (v) allow us to convert sound to other modalities when required.

Dinesh Nirmal of IBM starts with the question, "What is this guy smoking"? Speaking on making sense of regulations - that is long, unstructured, and complicated. He shares IBM Amelia , an NLP AI  then trained on regulatory documents and helps makes sense of it.

On Data Science Success (Hint: It gets boring)

Hillary Mason, GM of machine learning at Cloudera on "Practical ML". She asks, "How to go from data science as a niche practice to something that is more front and center delivering value?" And, "How much AI is embedded in your daily life?" It's boring because "it's inside", industrialized and hiding all the complexity. Fading into the background is what success feels like - so what does great look like? (i) Making live recommendations during call center interactions to improve the experience (saves money, improves experience (ii) Things that were expensive that become cheap; robo-advising, collaborative filtering of news  (new products/revenue). Hillary stresses taking experiments and scaling them; products->portfolios, people->collaborative organizations; technology scaling; a few data experts -> data organizations.

Big Data becomes Fast Data

Ziya Ma of Intel believes we are moving from Big Data to Fast Data - ie, real time data processing. Use cases are fraud detection, online shopping, and autonomous driving - feasibility and value driven by speed of data, analytics, and decisions. Intel believe feed driven by improving persistent memory improving capacity and cost. Here is the Intel stack and breakthroughs:

Julia Angwin is on stage to speak about "quantifying forgiveness" and got her "IT" start on an Apple II. And yes, she loved newspapers! Her first example is personal; journalism that led to people going to prison. But in data ... she investigated the algorithm that predicts whether you will do a crime (ie, real life Minority Report). The algorithm "looked like it may be biased." Her second example was on algorithms that predict car insurance prices based on zip codes that are weighted for locations where there is "more accident risks and costs."  She reports, "One groups is getting a break." Julia is launching a new newsroom investigating these types of questions.

Strata Data Awards!

Here are some awards announced at Strata Data NY:

• Most disruptive startup - TigerGraph
• Most Innovative Product - Dow Jones
• Most Impactful - Vertica

Smart Cities -> Smarter People

Chad Jennings of Google ends with our favorite topic; commuting traffic. Smart cities can't solve commuting and traffic problems by pouring concrete. His demo of "MyGeoTab" shows a heat map of traffic and then a predictive model on traffic based on temperature, visibility, precipitation, and traffic volume. Sadly, his demo was compromised by a slow internet connection.

Then Amanda Pustilnik spoke on brain-computer interfaces, ie, "The internet of the head." Research is on repairing and augmenting brain functions in patients with parkinson's disease and other conditions, but commercial augmentation use cases aim to help people become "better, faster, smarter." Amanda speaks to the many legal and privacy challenges with these technologies, What are the benefits vs risks  and who owns the data? What if governments or enterprises require brain sensors?

Ben Sharma says, "When you make information available from a variety of sources, no ideas emerges and the paradigms shifts." Static predictability is gone. His recommendations; "automate the knowns" and the repeatable processes, then "augment the unknowns" and the what ifs. "Be as disruptive as the disruptors" is his advice for banks and other industries, and the solution is data democratization.

Jacob Ward of CNN, Al Jazeera, and PBS lists decisions under uncertainty; "I know the type", "Happens all the time, I saw it on the news", and "This just feels like a real bargain" from seven papers published by Kahneman and Tversky 1971-1979.

Lots of questions... lots of thoughts.