Curious Data

What if the smartest thing you can do with data is to admit how much of it you have to build yourself?Diego Vallarino has spent a career at the sharp edge of risk, from Equifax and Scotiabank to Coface, where he and Bob first met, and now as a counselor to the board of the Inter-American Development Bank. He is also a prolific researcher who treats curiosity like a daily discipline.In this conversation, Diego breaks down how he builds models on open and synthetic data, why the source of your data matters far less for public policy than for a credit score, and how you can turn a thousand real observations into ten million and still trust the result. Then it takes a turn. He walks through his study of why people abandon a Netflix series, built with the very same survival model he once used on startups.Running through all of it is one human advantage no machine can copy. Listen now.Timestamped Highlights[00:02:02] The board seat that puts one economist in the room with 47 countries[00:04:58] From a soccer ball in Uruguay to econometrics, and the pandemic model that changed everything[00:08:32] Why economists quietly resist open data, and what finally changed their minds[00:10:24] Clean government data you can’t trust versus messy open data you can: which risk is worse[00:16:08] Synthetic data, explained: turning 1,000 real records into 10 million without breaking the model[00:19:52] Designing a public policy and testing it before you ever roll it out[00:24:53] Porándu: the Guaraní word for wonder that Bob just calls curiosity[00:28:34] The startup metric everyone asks about that has almost nothing to do with survival[00:36:38] The same risk model, now aimed at why you quit a Netflix series halfway through[00:47:29] The question that isn’t in any book, and why only humans can ask itMentioned ResourcesInter-American Development Bank (IDB)CofaceEquifax and ScotiabankAbout the GuestDiego Vallarino is a Uruguayan quantitative researcher and Counselor to the Board of Directors at the Inter-American Development Bank (IDB) and IDB Invest in Washington, D.C. He leads Sandpile Economics, an independent research program that applies self-organized criticality, network geometry, and complexity economics to understand how financial, productive, and institutional systems become fragile, and how small shocks turn into large crises. His work has been published or is forthcoming in journals such as Computational Economics, the Journal of Complex Networks, and Nonlinear Science. Before joining the IDB, he spent roughly a decade in risk management at Equifax, Scotiabank, and Coface, working with hundreds of Chief Risk Officers on how organizations measure and reduce risk.🔗 LinkedInImportant LinksConnect with Bob. Website | LinkedInBook a Strategy Session with Bob. Book a sessionLearn More About What We Do. What we doCurious Data is an original production hosted by Bob Lytle. Brought to you by Centri Analytics. Production and editing by Podcast Your Brand.

What if the polls everyone loves to distrust are actually more accurate than ever, and the real problem is how we read them?Collin Corbett has recruited and trained more than 3,500 candidates and helped put over 850 of them on the ballot. This year, he stopped running other people's campaigns and started his own, as an independent candidate for Illinois governor.Bob and Collin get into how modern polling really works, why weighting and modeling can quietly distort the numbers, why AI still cannot replace real data from real people, and what two decades of trial and error taught Collin about asking the right question.If you have ever wondered whether a poll can be trusted, or what it takes to turn messy public opinion into a smart decision, this one is for you. Listen now.Timestamped Highlights[00:02:08] The stretch of a campaign nobody sees coming, and why candidates are exhausted before it even starts[00:03:50] Why the moment you are needed is usually the moment you are least ready[00:07:31] How almost 38,000 signatures beat the you need 50,000 conventional wisdom[00:08:13] The data wizard named Abe who built their ballot system over a weekend[00:12:22] The rule Collin teaches every candidate about talking to voters[00:14:13] Can you trust a pollster? The first thing to check[00:18:08] Why asking AI how people feel gives you the wrong answer[00:19:09] Actual data versus model data, and why a model gets his own wife wrong[00:22:20] The same question, worded two ways, and the swing that follows[00:25:19] Running on heart or running on polls, and how the best do both[00:27:23] Illinois ranked 46th in growth, and the crony capitalism behind it[00:41:43] The book that priced out every way to win a single vote[00:45:41] The three phases of a campaign and the method that fits eachMentioned ResourcesGet Out the Vote: How to Increase Voter Turnout by Donald P. Green and Alan S. GerberAbout the GuestCollin Corbett is a political strategist and pollster with nearly two decades in politics and the founder of Cor Strategies, a political consulting firm based in Palatine, Illinois. He has recruited and trained more than 3,500 candidates and activists and helped hundreds of them reach the ballot. In 2026, he left the Republican Party and launched an independent campaign for governor of Illinois, running alongside Carolyn Schofield.🔗 LinkedIn | WebsiteImportant LinksConnect with Bob: Website | LinkedInBook a Strategy Session with Bob: Book a sessionLearn More About What We Do: What we doCurious Data is an original production hosted by Bob Lytle. Brought to you by Centri Analytics. Production and editing by Podcast Your Brand.

When a machine can pick the target, who keeps the humanity in the decision?Bob sits down with Brian Bengs, a retired military lawyer and independent U.S. Senate candidate in South Dakota, and Brianna Harter of the National Data Center Policy Project. The thread is not politics. It is who asks the questions when data and AI start making the calls.They get into keeping a human in the loop on life-and-death military decisions, why one proposed data center could draw the power of 400,000 South Dakota homes, and what happens to local water, power bills, and voice when big money moves faster than the people it affects.A clear-eyed look at curiosity, governance, and the facts underneath the headlines. Listen now.Timestamped Highlights[00:02:33] A graduating class of 17, and a Navy hitch that never left dry dock[00:04:09] How a four-year plan in the JAG Corps became two decades[00:09:37] Inside the room where a lawyer signs off before a strike[00:11:08] The one decision Brian says AI should never make alone[00:12:09] Why only humans can be curious, and machines just execute[00:14:34] The war power the Constitution hands Congress, and almost no one uses[00:17:12] The link between a distant war and the price of your groceries[00:20:53] 400,000 homes of power in a state with 366,000 homes[00:22:25] Follow the money behind South Dakota’s data center push[00:26:58] What data centers will not tell you about their water and power[00:31:18] Why 45% of voters are independents, and both parties want it quietMentioned ResourcesBengs for South DakotaNational Data Center Policy Project, the data center policy reform effort directed by Brianna Harter, focused on Virginia and the federal level.Veterans for All VotersAbout the GuestsBrian Bengs. Retired Lieutenant Colonel Brian Bengs is a 26-year veteran of the U.S. Navy and U.S. Air Force. In the Air Force Judge Advocate General's (JAG) Corps, he worked as a federal ethics advisor and prosecutor, holding Air Force personnel to the branch's core values. He has taught law and policy at the U.S. Air Force Academy, the NATO School in Germany, and Northern State University in Aberdeen, South Dakota, and has served as a National Park Ranger. A registered independent for most of his adult life, Brian is now running for U.S. Senate in South Dakota.🔗 WebsiteBrianna Harter. Brianna Harter is a leader in Industrial-Organizational Psychology who helps organizations navigate change management and responsible AI adoption. As director of the National Data Center Policy Project, a policy reform effort focused on data centers in Virginia and at the federal level, she advocates for transparent data and policies that help citizens make informed decisions about data center development in their communities. She champions AI with meaningful human oversight, keeping people at the center of how AI shapes work and life.🔗 LinkedInImportant LinksConnect with Bob: Website | LinkedInBook a Strategy Session with Bob: Book a sessionLearn More About What We Do: What we doCurious Data is an original production hosted by Bob Lytle. Brought to you by Centri Analytics. Production and editing by Podcast Your Brand.

When anyone can buy the same AI, what still gives one side the edge?Steve Woll spent the last eighteen years of his Navy career turning weather and ocean data into decisions that ships and lives depended on. Now a consultant and an independent candidate for Congress in Virginia, he joins Bob to work through how data actually creates advantage, and where that advantage is quietly shifting.They get into the split between physics-based forecasting and AI pattern recognition, the post-war pact that makes even rivals share their weather data, why a military model cannot afford to hallucinate the way a chatbot can, and the one problem worse than bad data: data that goes missing.It is really a conversation about where the human edge goes once the machines catch up. Listen now.Timestamped Highlights[00:03:27] A zoology and computer science major at Duke, decades before anyone called it data science.[00:10:39] What a Navy meteorology officer actually works with, and why it all comes back to physics.[00:12:15] The hardest part of forecasting: merging a dozen kinds of data into one picture.[00:13:15] The post-war pact that makes even rivals hand over their weather data.[00:14:49] Why the military shares some of its data openly and quietly keeps the rest.[00:16:41] Cray supercomputers, and who really had the best forecasting on the planet.[00:18:00] Two forecasting systems now run side by side: physics equations and AI pattern recognition.[00:19:16] Why a weather model cannot afford to hallucinate the way a chatbot can.[00:21:23] When anyone can buy the same AI, what happens to a nation’s data edge?[00:30:41] The worst thing for a data analyst is not bad data. It is missing data.[00:49:13] Why human curiosity may hold the last edge over the machines.Mentioned ResourcesWorld Meteorological Organization (WMO)National Oceanic and Atmospheric Administration (NOAA)Ocean Observatories Initiative (OOI)Internet ArchiveBlockers: Rebels in the Deep State by Michael LewisFederal Election Commission (FEC)About the GuestSteve Woll is a retired U.S. Navy officer who finished his career as a Commander. He was surface warfare qualified and spent the last eighteen years in uniform as a Special Duty Officer for Meteorology and Oceanography, turning weather and ocean data into operational decisions. Raised in Pittsburgh, he studied zoology and computer science at Duke University and began his career in environmental science computer modeling before joining the Navy through Officer Candidate School. Today, he consults with data-intensive organizations, has worked extensively with NOAA, and is running for U.S. Congress as an independent in Virginia's 3rd District.🔗 LinkedIn | WebsiteImportant LinksConnect with Bob: Website | LinkedInBook a Strategy Session with Bob: Book a sessionLearn More About What We Do: What we doCurious Data is an original production hosted by Bob Lytle. Brought to you by Centri Analytics. Production and editing by Podcast Your Brand.

What if the biggest source of AI waste is not the data centers, but the rest of us?Bob sits down with AI architect, technologist, and ethicist Matt Konwiser for a conversation about the gap between how loudly we argue about AI and how little we actually understand it.They get into why governance keeps falling behind, what the real water and energy numbers behind a data center actually say, and why the fights over slop, disinformation, and trust may come down to individual choices more than policy.It is a first-principles look at using a powerful technology wisely, from someone who builds with it every day. Listen now.Timestamped Highlights[00:04:53] Thirty years in tech, and the throughline that led Matt to AI governance.[00:06:16] Why Matt says almost no company is truly dealing with AI governance.[00:07:25] The schoolyard speed limit that explains what ethics really is.[00:09:12] How Matt turns a daily news habit into a public voice, with nothing scheduled.[00:11:57] The zero-trust lesson from cybersecurity.[00:13:23] AGI and RSI in plain terms, and why together they are new territory.[00:16:26] The February to September swing that split two allies.[00:19:51] Is everything AI now, and have people stopped caring?[00:24:43] The line between useful AI and the slop quietly draining resources.[00:37:13] The water math behind a single data center, and what the headline number hides.[00:42:11] Jensen Huang’s blunt rule for anyone building something that is not safe.[00:46:11] What if the smartest AI simply refused to do our wasteful work?[00:51:27] Matt’s closing case for hope about AI and humanity.Mentioned ResourcesEnvironmental Cost of Artificial Intelligence: Carbon, Water and Land FootprintsAbout the GuestMatt Konwiser is a Field CTO at IBM and a longtime technologist whose work spans telecommunications, cybersecurity, data, and artificial intelligence. Over three decades he has advised executive teams on how to actually put new technology to work, and today he is one of the more direct public voices on AI governance and ethics. He also serves on the AI Advisory Council of the Global Technology Industry Association, advises a carbon-first marketing startup, and co-hosts the AI Diatribe podcast. The views he shares here are his own.🔗 LinkedIn | AI DiatribeImportant LinksConnect with Bob: Website | LinkedInBook a Strategy Session with Bob: Book a sessionLearn More About What We Do: What we doCurious Data is an original production hosted by Bob Lytle. Brought to you by Centri Analytics. Production and editing by Podcast Your Brand.