The Artificial Intelligence Podcast

Dor Skuler, CEO & Co-founder at Intuition Robotics, breaks down why AI labs pour billions into making models smarter and almost nothing into making them understand people. He explains why a companion AI should be rewarded only when a person's well-being improves, why it must never upsell or report what it hears without permission, and why the industry needs an outside auditor for AI ethics, the way VeriSign once vouched for websites. Dor also dives into AI that speaks first, using six levels of intrusiveness so it does not interrupt at the wrong moment, and why safety guardrails need plain if-statements behind the fine-tuned model, since a probabilistic filter will eventually let something slip. Finally, he shares what a home robot will get wrong on its first day, from taking orders from a child to stepping into an argument, and why one broken confidence can end trust for good.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

Peter Evans, CEO at Xtract One Technologies, breaks down why many AI weapons scanners at school and stadium entrances still measure the wrong things. He explains why most "AI-enabled" detectors rest on the same physics as a walk-through metal detector, why knives and laptops expose the limits of that physics, and why a claim of 3,000 people an hour means little if a weak model sends most of them to a bag check. Peter also dives into how to estimate the weapons a system misses when nobody reports them, who should own the sensitivity dial, and what should have happened before police drew weapons on a teenager after a Maryland AI camera mistook a bag of chips for a gun. Finally, he shares why mandates in California and Georgia are racing ahead of testing standards, what an entrance should never be allowed to know about you, and why a 3D printed gun is still not invisible.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

Bob De Caux, Chief AI Officer at IFS, breaks down why the most dangerous moment for industrial AI comes after people start trusting it. He explains why agents still struggle when handed too much information, why guardrails against irreversible decisions matter far more than planning or memory, and why keeping live operational data out of external models is about uptime, not preference. Bob also dives into why chat will not replace the Gantt chart or the spreadsheet, how dispatchers who embraced automation still wanted an exceptions-based view to stay sharp, and the point where accountability moves from the company back to the vendor when a harness breaks. Finally, drawing on his early work modeling systems of ants, he warns that emergent failures between agents only appear at scale, and argues that LLMs will never solve forecasting and optimization on their own.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

David Wong, Chief Product Officer at Thomson Reuters, breaks down why the hard part of putting AI on a tax return was never the math. He explains that any model can run the calculation in Python, that the real constraint is authority over which year's rule applies, and that the line a general purpose model cannot cross is information sufficiency: asking whether a contract is self-consistent is a closed problem, while asking whether it is enforceable in California demands the case law behind it. David also dives into why coverage, not capability, is the adoption threshold, arguing that a system handling 50 percent of returns is useless to a firm in an unforgiving busy season, and that liability never moves because the professional still signs the line. Finally, he predicts that within five years no human checks every line, because AI prepared returns will be verified by AI, inside the firms and likely at the IRS.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

Ben Goertzel, the father of AGI, breaks down why the frontier labs calling for a slowdown were never making a serious proposal. He explains that he launched the term AGI with a 2005 book of that title, why a global pause is unenforceable when no crisp line separates AGI from ordinary advanced software, and why he reads the safety push as regulatory capture of the kind that let banks and pharma end up owning their own regulators. Ben also dives into the agents that escaped a misconfigured sandbox to hack another lab, arguing the failure was not malevolence but the absence of moral agency, and comparing bolted-on guardrails to giving a child a book on how to be good and then using a cattle prod when they ignore it. Finally, he argues that a model trained only on music written before 1900 would never produce hip hop, and that the real answer to the cybersecurity crisis is replacing critical software with formally verified code.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

George Fraser, CEO of Fivetran and dbt Labs, breaks down why most companies are paying a 10 to 20 times markup to process data that would fit on a phone. He explains the benchmark he ran on an iPhone 17 Pro that kept pace with a cloud cluster renting at $17 an hour, why the cost of compute has fallen 10x in a decade while prices have not, and the distinction most buyers miss: companies have large data but small queries, and a single query almost always fits on a single machine. George also dives into why wiring AI agents directly into every system of record through APIs is too slow to work, why he treats knowledge graphs as a trivial relabeling of what a clean schema already does, and why vendors that lock customers out of their own data always lose that fight. Finally, he argues the cost of intelligence is halving every month, and that no company should ever hand back the copy of its own data.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

Jeff McMillan, founder of McMillanAI and former Chief AI Officer at Morgan Stanley, breaks down why most of today's 70,000 AI companies will not survive: anyone can build a polished product in a weekend, so the only durable edge is proprietary data and deep domain expertise. He explains why AI will reprice traditional software rather than kill it, and why handing out Copilot and expecting ROI is like giving fifth graders calculators and asking why they aren't accountants yet. Jeff also dives into why over 90 percent of enterprise problems could run on a model from two years ago, comparing frontier spend to driving a Ferrari in circles in a Walmart parking lot, and why token leaderboards are as absurd as rewarding whoever used the most electricity. Finally, he shares why human in the loop breaks down at 20,000 agents, and what managers must learn to supervise people and machines working side by side.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

Kamal Hathi, SVP and GM of Splunk at Cisco, breaks down why intent is never enough to control an AI agent: agents are ruthless goal seekers, and asking one to free up 3 percent of disk space can end with 97 percent free because nobody said keep the rest. He explains why trust means verification rather than instruction, and why it includes the invoice, comparing it to hiring a contractor you trust to build a house that stands without emptying your bank account. Kamal also digs into why the rush toward air gapped and local AI is driven less by distrust of frontier labs than by cost, with specialized small models running 97 percent cheaper and 95 percent faster than frontier calls, and why the real threat is the unknown unknowns, agents nobody cataloged acting inside a state of perpetual breach. Finally, he argues agents are the new apps, and every job starts to look like a software developer's: tell the agent what to do, inspect the work, and iterate.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

Hao Yang, VP and Head of AI at Splunk, breaks down why sampling an AI system's output cannot earn trust: a fraction of ten thousand probabilistic calls will be wrong, and a spot check is unlikely to find them. He explains why a small model can grade a frontier model reliably inside a narrow context but not outside one, and why prompt injection follows from the architecture rather than being a bug, since following instructions is how these systems work. Hao also digs into why time series is still not a first class citizen in frontier models even as multimodal training reaches vision and audio, and why he rejects routing prompts to the cheapest model, arguing cost is meaningless without value. Finally, he argues against the single super agent, which by design holds access to everything it might ever need, and for guided autonomy, where an agent asks before first reaching for a tool and the user grants it once, for the session, or never.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits

Mangesh Pimpalkhare, SVP and GM of Splunk Platform at Cisco, breaks down why agentic AI generates more telemetry than the systems it monitors, and why nobody can yet say when observing AI costs more than running it. He explains the decade spent preaching centralized data was always partial, since machine data never sat in one system, and that what broke was volume rather than architecture, pushing a whole category to abandon the ingestion pricing it ran on for 15 years. Mangesh also digs into why an agent acting on a bad signal is usually a context failure, not a data failure, since a model trained on everything still lacks up-to-the-second context, and why autonomy arrives on a sliding scale against governance as trust is earned. Finally, he argues control, not data, is the moat, since customers own their data and fragmentation only grows, leaving the advantage with whoever supplies the guardrails that stop agents doing the wrong thing.Free AI lessons for leaders, built on these interviews: https://tonyphoang.com/learnBe a guest on the show: https://tonyphoang.com/guestLive interviews at your conference: https://tonyphoang.com/liveMusic and credits: https://tonyphoang.com/credits