AI-Curious with Jeff Wilser

Could AI agents become better traders than humans—and what happens when “decision-making” gets outsourced to software that can act at machine speed?In this conversation, we go deep with Saad Naja, founder of PIP World, on the rise of AI agent auto-traders: multi-agent “swarms” that resemble a miniature trading desk—specialist analysts feeding into an AI “portfolio manager” that can decide whether to buy, sell, or hold. Even if you’ve never day traded, finance may be one of the clearest real-world testbeds for autonomous agents—because markets keep score in real time.Key moments[00:02:00] How AI has quietly shaped trading for decades—long before ChatGPT[00:05:00] Why retail traders lose so consistently: data disadvantage + execution problems[00:10:00] What’s changed with generative AI: analysis that used to take teams can now happen fast[00:12:00] Why “AI swarms” differ from old-school trading bots (context, coordination, and specialization)[00:17:00] The “trading desk in software” model: specialist agents + a chief decision-maker[00:21:00] How PIP World trained and tested models—and why win-rate isn’t the whole story[00:26:00] Why they launched in simulation first—and what it reveals about performance[00:30:00] How agents trade differently than humans (patience, confirmation, discipline)[00:37:00] Hallucinations, guardrails, and why specialization reduces “AI going rogue” risk[00:40:00] The endgame: “agent vs. agent” markets, shrinking edges, and the data arms race[00:45:00] A 5-year prediction: how much trading could become fully agentic[00:47:00] Why crypto/DeFi is a natural early proving ground—and how TradFi could followWhat you’ll hear us exploreThe difference between traditional algo trading (single-strategy rule sets) and agentic systems (multiple specialized “analysts” + a coordinating decision layer)Why most retail traders aren’t necessarily wrong on ideas—but lose on execution and risk managementHow “edge” shifts when everyone has access to powerful models: data quality, workflows, and strategy selectionWhat finance teaches us about the broader economy as agents move from “assistants” to “actors”If you’re curious about autonomous agents—whether you trade or not—this is a concrete, high-stakes preview of what “agentic work” could look like when the scoreboard is real.Guest: Saad Naja, Founder, PIP WorldTopics: AI agents, multi-agent swarms, algorithmic trading, market data, risk management, DeFi, agentic automation

Can AI actually help eradicate poverty for real people, right now—not in some vague future?We talk with two leaders from Opportunity International who are trying to do exactly that, using AI to support smallholder farmers and low-cost private schools across Africa and beyond.In this episode of AI-Curious, we sit down with Ama Akuamoah and Paul Essene from Opportunity International’s Digital Innovation Group. We explore how they’re deploying AI chatbots over WhatsApp to help farmers diagnose crop diseases, optimize planting decisions, and access localized agricultural advice, and how they’re building classroom tools that give overstretched teachers better lesson plans and more time for their students.We hear the origin story of their farmer chatbot—from a mud-brick home in Malawi to pilots now running in five countries—and the 80-year-old farmer who saved her okra crop by using an AI tool through a trusted “farmer support agent.” We also dig into how they use retrieval-augmented generation (RAG) grounded in local government content, why “human in the loop” is non-negotiable, and what it really takes to make AI work in communities with limited electricity, spotty connectivity, and low digital literacy.Along the way, we talk about ethics and trust: data consent, privacy for highly vulnerable populations, and the risk of leaving people behind in this new wave of AI. And we zoom out to the bigger picture—why conversational AI in local languages could be a genuine game-changer for economic development if infrastructure, funding, and partnerships keep pace.What we cover[01:00] Opportunity International’s mission and why they focus on farmers, teachers, and micro-entrepreneurs[08:00] The Malawi farm-floor moment that sparked their AI journey[09:00] How a WhatsApp-based chatbot helps thousands of farmers, and how “farmer support agents” multiply its impact[13:40] Using RAG and local government content to keep answers accurate and context-aware[15:30] Bringing AI into crowded, low-resource classrooms and supporting teachers with lesson plans and copilots[20:15] The hard parts: infrastructure gaps, low-cost devices, digital literacy, and why this work is heavy lifting[24:30] Human-centered design in action: co-creating with communities, iterating in the field, and learning from pilots[37:50] Guardrails, consent, and building trust around AI in vulnerable communities[41:00] What’s needed for real scale: infrastructure, funding, language support, and the right partners[43:00] Their hopeful vision for AI as a lever for economic development—if no one gets left behindIf you’re interested in AI for social impact, global development, or what it really takes to deploy AI outside Silicon Valley, this conversation is a grounded, hopeful look at what’s already working—and what still needs to change.

Is AI making us smarter or dumber—and how do we make sure we’re on the right side of that divide?In this episode of AI-Curious, we talk with Professor Vasant Dhar, author of the new book Thinking With Machines: The Brave New World of AI. Vasant isn’t just a historian of AI; he’s part of the story. In the 1990s, he helped bring machine learning to Wall Street, founded one of the world’s first ML-based hedge funds, and became the first professor to teach AI at NYU Stern, where he’s now the Robert A. Miller Professor of Business. He also hosts the podcast Brave New World.We explore how AI evolved from early efforts around “thinking, planning, and reasoning” to the long era of pure prediction and machine learning, and then to today’s general-purpose models that blur the line between expertise and common sense. Vasant explains why the autocomplete problem turned out to be a gateway to something like “general intelligence,” and why that matters for how we define knowledge, understanding, and reasoning.We then dive into finance and the search for “edge.” Vasant shares war stories from his days at Morgan Stanley, where machine learning systems quietly reshaped trading strategies and risk-taking. We unpack his work on “the DaBot,” an AI built on the writings and valuation framework of Aswath Damodaran, and what happens when every analyst and firm can tap this kind of supercharged valuation machine. Does AI erase the edge—or simply raise the bar for everyone?Finally, we zoom out to careers, education, and everyday life. Vasant argues that AI is likely to bifurcate humanity into those who become “superhuman” by thinking with machines, and those who outsource their thinking and fall behind. We discuss how classrooms will change, why many teachers (and professors) may be more automatable than they realize, and how each of us can periodically test whether AI is making us smarter or dumber.If you’re curious about how to work with AI rather than be replaced or outpaced by it, this conversation offers a grounded, big-picture way to think about your edge in the age of intelligent machines.

Can a city use AI to cut red tape, fill potholes faster, and shave minutes off commutes—without sliding into surveillance? We sit down with San José’s mayor, Matt Mahan, to unpack how a highly regulated public institution can adopt AI pragmatically and responsibly. In this episode, we dig into the playbook: pilots that become policy, guardrails that build trust, and workforce upskilling that actually moves the needle.We cover how bus routes now hit fewer red lights, why real-time translation boosts civic inclusion, what “privacy by design” looks like for license-plate readers, and how a 10-week AI curriculum is turning city staff into hands-on builders. We also press on the risks—bias, privacy, and transparency—and explore where city AI is headed next: transit, permitting, and procurement.HighlightsFrom pilots to scale: Bus route optimization with Light AI cut red-light hits by 50%+ and reduced travel time by 20%+, now rolling out citywide.Inclusion by default: Real-time multilingual access (e.g., Wordly) and improved translations informed by San José’s deep Vietnamese-language data.Eyes on the street, not faces: No facial recognition, strict retention, no third-party data sharing, and tightly controlled access to ALPR data.Upskilling at scale: A 10-week AI curriculum (plus a data track) with San José State; staff build custom GPTs (including a budget-analysis GPT) to speed analysis.Culture that ships: A “coalition of the willing,” clear problem statements, and a Mayor’s Office of Technology & Innovation to operationalize change.Road ahead: Smarter mass transit, faster permitting, and streamlined procurement—practical abundance without new tax dollars.If you’re new here, we’d love your support—subscribe on Apple, Spotify, or YouTube, and consider leaving a quick rating or sharing this episode with a colleague who’s wrestling with real-world AI adoption.

Does AI make us more creative—or quietly replace us?In this episode of AI-Curious, we sit down with Dr. Maya Ackerman—author of Creative Machines: AI, Art, and Us—to probe where human creativity ends and machine creativity begins, and how incentives in Big Tech and venture capital shape the tools we all use. We explore why today’s dominant systems skew “convergent” (safe, samey, oracle-like) instead of “divergent” (surprising, generative), what that means for artists, and how to design AI that actually elevates human imagination rather than displacing it.Why listenWe wrestle with uncomfortable truths: bias mirrored back at us, investor pressure to “replace” vs. “augment,” and the risk of a cultural sea of slop. We also map a constructive path forward—collaborative systems, richer human–AI interfaces, and a 10-year horizon where AI expands human creative range.GuestDr. Maya Ackerman — AI researcher, entrepreneur, and author of Creative Machines: AI, Art, and Us. TakeawaysAI reflects us. Bias in → bias out; representation fixes are not enough without cultural understanding.Incentives matter. Many well-funded tools are architected to replace creators; augmentation tools are underfunded.Creativity ≠ autocomplete. Today’s LLMs are optimized for correctness and convergence, not genuine divergence.Better interfaces beat bigger models. Beyond “text-to-X,” human-centred, interactive tools can coach, not usurp.A hopeful arc. With the right design, collaborative AI can measurably raise human creative ability—and stick.Dr. Ackerman's new book: Creative Machineshttps://www.amazon.com/Creative-Machines-Future-Human-Creativity/dp/1394316267

Deepak Agarwal, LinkedIn's Chief AI Officer, discusses the integration of AI in hiring and job searching. He shares insights on how AI agents, like the Hiring Assistant, streamline recruitment and free up time for human connection. Deepak emphasizes the importance of responsible AI practices, including bias detection and governance. He also highlights the shift from keyword searches to semantic job searches for better matches. His vision for the future includes AI enhancing human creativity and making the hiring process more effective and authentic.

In this discussion with Curtis Sparrer, co-founder of Bospar PR and president of the San Francisco Press Club, they explore the shift from classic SEO to Generative Engine Optimization (GEO). Curtis reveals how AI is reshaping brand visibility and emphasizes the importance of reputable sources in this new landscape. He shares insights on maintaining effective SEO tactics while adapting to AI-first strategies. Curtis also discusses the risks of misinformation, the changing dynamics in pitching, and the need for companies to embrace proactive approaches to succeed in the AI era.

What happens when “space robots” stop being sci-fi set dressing and start punching a clock? We dig into a new breed of microgravity robots that do the unglamorous work—so astronauts can do more science.In this episode of AI-Curious, we talk with Ethan Barajas (CEO) and Jamie Palmer (CTO), co-founders of Icarus Robots, fresh out of stealth with a $6M raise. Their pitch is simple and radical: put agile, teleoperated robots insidespacecraft like the ISS to handle cargo, inspections, and maintenance—then use the resulting microgravity manipulation data to unlock partial (and eventually full) autonomy. We cover the tech, the economics (why astronaut time is so expensive), the AI roadmap, and a pragmatic path from today’s chores to tomorrow’s orbital factories and lunar bases.What we coverWhy astronaut hours are precious—and how robots can “augment” rather than replace themThe form factor: free-flying, drone-like bodies with dual arms optimized for zero-G dexterityInside first, outside later: a deployment strategy that lowers safety hurdles and accelerates learningData advantage: building the first large microgravity manipulation dataset via continuous teleopAI’s role: from human-in-the-loop control to primitives to scalable dexterous manipulationCommunications and latency: S-band today, laser links tomorrow; what “real-time” actually meansThe “orbital factory” thesis: pharma, semiconductors, fiber optics—and servicing orbital data centersLong-horizon forecasts: humans living and working in space; physical labor increasingly done by robotsGuestsEthan Barajas — Co-founder & CEO, Icarus RobotsJamie Palmer — Co-founder & CTO, Icarus RobotsWhy this mattersIf half of Earth’s GDP is labor, the space economy scales only when on-orbit labor scales. Teleoperated robots that learn from expert demonstrations—then graduate to safe autonomy—are a credible bridge from today’s stations to tomorrow’s factories, data centers, and off-world bases.https://www.icarusrobotics.com/

What if “AI teammates” aren’t sci-fi at all, but the next mundane tool that quietly kills Monday dread?In this episode of AI-Curious, we sit down with David Shim, CEO of Read.ai, to unpack what workers actually want from AI, how teams are adopting agents from the bottom up, and what a practical “digital twin” might do at work—minus the Black Mirror vibes. We cover fast-path ROI (meeting notes → action items), the shift from “prompts” to ambient workflows, and why the most valuable corporate asset may soon be the storage of intelligence—the living record of how your organization thinks and decides.What we coverWhy 70% of workers say they want AI agents—and what basic tasks deliver real ROI nowA crawl-walk-run roadmap: note-taking → briefing → follow-ups → lightweight agents → digital twin“Storage of intelligence” as a competitive moat (institutional knowledge that doesn’t walk out the door)Guardrails, data separation, and how to make privacy concerns non-negotiableBottom-up adoption: why employees are forcing IT’s hand—and how leaders should respondThe macro view: augmentation vs. replacement, and the provocative idea that AI replaces computers (as the interface)If you find this useful, we’d love a rating and a quick share with a teammate who’s piloting AI at work.Read.AI:https://www.read.ai/

AI is often framed as a climate problem—energy-hungry data centers, ballooning carbon emissions, and talk of nuclear power just to keep the servers running. But could AI also become part of the solution?In this episode of AI-Curious, we sit down with Josh Dorfman—climate tech entrepreneur and host of Supercool—to explore how artificial intelligence might help tackle climate change. Josh doesn’t offer hand-wavy promises. Instead, we dive into concrete examples where AI is already making a difference.What we cover:[4:17] Josh’s background at the intersection of technology, climate, and business.[8:18] How AI data centers are impacting energy use—and why fossil fuels can’t scale to meet demand.[12:30] The role of nuclear, geothermal, and solar-plus-storage in powering AI sustainably.[23:25] AI-optimized school buses: how Oakland electrified its fleet with fewer vehicles.[27:44] BrainBox AI and smarter buildings: cutting emissions through predictive HVAC optimization.[31:42] AI in waste management: from pneumatic trash tubes to AI sorting recyclables.[41:17] Big-picture futures: AI efficiency, plummeting solar costs, and the possibility of “trivially cheap” energy.The conversation blends realism with optimism—grounded in the challenges of energy demand, yet hopeful about AI-driven solutions in transportation, buildings, waste, and renewable power.If you’ve ever wondered whether AI can be more than an energy drain—and instead help drive sustainability—this episode offers both perspective and inspiration.🎧 Subscribe to AI-Curious:• Apple Podcastshttps://podcasts.apple.com/us/podcast/ai-curious-with-jeff-wilser/id1703130308• Spotifyhttps://open.spotify.com/show/70a9Xbhu5XQ47YOgVTE44Q?si=c31e2c02d8b64f1b• YouTubehttps://www.youtube.com/@jeffwilser