Podcast
Eye On A.I.
Hosted by Craig S. Smith · EN
Eye on A.I. is a biweekly podcast, hosted by longtime New York Times correspondent Craig S. Smith. In each episode, Craig will talk to people making a difference in artificial intelligence. The podcast aims to put incremental advances into a broader context and consider the global implications of the developing technology. AI is about to change your world, so pay attention.
51episodes
Episodes
Newest firstAll episodes
Training AI Models Without a Billion-Dollar Data Center | Steffen Cruz of Macrocosmos
3d ago00:47:11Tap to summarizeTraining a frontier AI model today requires hundreds of thousands of GPUs, months of compute time, and a budget that only a handful of companies on earth can afford. Steffen Cruz, co-founder and CTO of Macrocosmos, thinks that model is about to break, and he's spending his time building what comes next. His project IOTA, operating within the BitTensor blockchain ecosystem, uses distributed training to split large language models across thousands of devices located around the world, coordinated by blockchain, and powered by surplus cheap energy wherever it exists. After nine months of research, the system can reproduce baseline benchmark performance using what Cruz calls "wonky vegetables" - unreliable, churning, globally distributed compute - and turn it into something indistinguishable from centralized training if you use the right approach. The conversation with Craig Smith covers the mechanics of how this actually works, why the blockchain's role is far narrower and more practical than most people assume, and why the Mac mini stockpiling trend creates an unexpected supply of distributed compute that can earn passive income when idle. Cruz's target: a 70 billion parameter model by mid-2025, trained at 10-20% of what it would cost through a hyperscaler, and aimed squarely at the legal firms, hospitals, and cash-strapped startups that have been waiting to train their own sovereign models but couldn't afford the price tag. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.
Transcribe →
The Single Biggest Barrier to AI Adoption Isn't the Technology — It's This | Errol Gardner of EY
6d ago00:54:59Tap to summarizeErrol Gardner has spent 35 years advising the world's largest organizations through major technology transitions, and his assessment of where enterprise agentic AI actually stands is one of the most grounded you'll hear anywhere. His number: less than 1 out of 10 on a maturity scale. Not because the technology isn't ready, but because deploying agentic AI across an organization doesn't tweak how it works, it requires rebuilding how it works. And that is a fundamentally different kind of challenge than anything the AI hype cycle is currently acknowledging. In this conversation with Craig Smith, Gardner walks through why cloud adoption still hasn't reached 7 out of 10, what that means for agentic AI timelines, why the single biggest barrier to adoption is human resistance rather than technical limitation, and why governments will ultimately have to step in to manage workforce displacement at scale. He also raises a question that almost nobody is asking: is the value exchange between the technology sector and traditional industries sustainable in the long run? It's a conversation that doesn't just describe where AI is, it explains why the gap between the narrative and the reality has never been wider. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.
Transcribe →
Oliver Dial of IBM: Quantum Advantage Is Happening This Year
1w ago00:50:55Tap to summarizeIBM's VP of Quantum Systems, Oliver Dial, has spent his career building quantum computers from the ground up, and he's unusually direct about what they can and can't do. In this conversation with Craig Smith, Oliver Dial walks through where the field actually stands in 2026: quantum utility was achieved in 2023, quantum advantage is the target for this year, and a fully error-corrected machine capable of tackling the hard problems is on IBM's roadmap for 2029. That last milestone, Dial says, now feels both achievable and terrifying. The episode is worth your time because Dial doesn't hype. He explains why IBM built a 1,000-qubit computer and then took it apart almost immediately, why Google's quantum advantage claims remain scientifically contested, and how a new error-correcting code IBM developed just reduced the qubit overhead required for fault-tolerant quantum computing by an order of magnitude. For anyone trying to understand what quantum computing will actually mean for their industry, and when, this is the clearest map of the road ahead available right now. If this conversation changed how you think about the future of computing, subscribe to Eye on A.I. for weekly conversations with the researchers and builders shaping what comes next.
Transcribe →
Why Agentic-First Startups Won't Disrupt Enterprises as Fast as Everyone Thinks | Kris Lovejoy
2w ago00:56:57Tap to summarizeKris Lovejoy, Global Strategy Leader at Kyndryl, has spent her career at the intersection of IT infrastructure and security. Right now, she's one of the people enterprises call when they want to move from AI experimentation to real deployment. Her diagnosis is clear: agentic AI is a bullet train sitting on tracks built for 30 miles per hour. The technology is ready. Most organizations aren't, and the gap between a successful pilot and a production system running at scale is far wider than the hype suggests. In this conversation with Craig Smith, Lovejoy walks through why IT service management is the smartest entry point for agentic adoption, how cost savings of up to 90% in that area can fund broader modernization, and why the security risks in agentic systems are less about sophisticated hackers and more about misconfiguration, bad context, and human error. She closes with a specific prediction: half of traditional IT administration tasks will be handled by AI agents by 2031, and a surprising take on who will actually thrive in the agentic era: not coders, but people trained to ask the right questions. For anyone making decisions about AI adoption, this is the most practical conversation available right now. Subscribe to Eye on A.I. for weekly conversations with the people building and deploying the future of AI.
Transcribe →
Loris Degioanni: Why AI Is Breaking Cybersecurity, and What Comes Next
3w ago00:51:15Tap to summarizeAI has fundamentally changed the cybersecurity threat landscape, not by inventing new attack types, but by collapsing the timeline. The same tools that make software developers more productive are now being used by attackers to move from vulnerability disclosure to active exploit in a matter of hours. That shift, argues Loris Degioanni, CTO and founder of Sysdig, changes everything about how defense needs to work. In this episode, Craig Smith talks with Loris Degioanni about why human-centered security is becoming a structural liability, what "headless cloud security" means in practice, and why the coding agent (tools like Claude Code or Codex) may become the new operating system through which all enterprise security workflows run. It's a conversation about architecture, urgency, and what it actually means to fight a tank when you've been trained to use a baseball bat. If this conversation made you think differently about AI and security, subscribe to Eye on A.I. for weekly conversations with the people building and defending the future.
Transcribe →#342 Andrew Thangaraj: The $5,000 IIT Degree: Can India Fix Its Broken Education System?
4w ago00:48:53Summary readyWhat if the most competitive exam in the world is also the most destructive? In this episode of Eye on AI, Craig Smith sits down with Professor Andrew Thangaraj, faculty at the Department of Electrical Engineering at IIT Madras, to explore how one of India's most prestigious institutions is quietly dismantling the system it helped build. Andrew lays out the honest reality of higher education in India. Two and a half crore kids reach college age every year. Only 90 lakh make it to college. And the IITs, the most coveted institutions in the country, take just 17,000. The competition to reach those seats has become so extreme that students are losing their childhoods, their development is stunted, and even those who make it through are often unemployable because the system rewards knowledge over skills. Andrew walks through exactly how IIT Madras is responding. A full, IIT-branded undergraduate degree in data science delivered entirely online for under five lakhs, roughly $5,000. No JEE required. No elite school background needed. Just a 10th standard foundation and the willingness to do the work. The program flips the traditional model, putting hands-on skills and real projects before theory, building in multiple exit points for students who need to start earning before they finish, and scaling to over 40,000 active students through a hybrid of faculty-recorded lectures, full-time instructors, and a remarkably active student community. We also get into the bigger picture. Why India's AI talent gap is as much a culture problem as a numbers problem. Whether India can leapfrog into AI leadership the way China did after rebuilding its research ecosystem. Where AI tools are already being tested inside the program and where they still fall short. And how AI deployed in Indian languages, in agriculture, and in the courts could drive the kind of societal change that no corporate productivity tool ever will. Subscribe for more conversations with the people shaping the future of AI and emerging technology. Stay Updated: Craig Smith on X: https://x.com/craigss Eye on A.I. on X: https://x.com/EyeOn_AI (00:00) Introduction and Andrew Thangaraj's Background (01:29) India's Higher Education Bottleneck (03:45) Designing a $5,000 IIT Degree (09:27) Why Graduates Still Lack Skills (12:31) When the Program Started and How It Got Approved (13:56) Program Structure, Diplomas and Multiple Exit Points (17:52) Who the Program Reaches and Surprising Student Stories (24:57) Older Students, Working Professionals and International Enrollment (29:55) Can India Leapfrog in AI (34:03) Data Centers, Power and Infrastructure Gaps (40:57) How Involved Are the IITs in India's AI Mission (46:00) AI for Languages, Farms and Courts
Summary#341 Celia Merzbacher: Beyond the Buzzword: The Real State of Quantum Computing, Sensing, and AI in 2025
4w ago00:44:55Tap to summarizeWhat does the quantum industry actually look like right now, beneath all the hype? In this episode of Eye on AI, Craig Smith sits down with Celia Merzbacher, Executive Director of the Quantum Economic Development Consortium (QED-C), to break down the real state of quantum technology in 2025. From market growth and enterprise readiness to the growing intersection with AI, Celia brings a grounded insider perspective on where the industry stands and what comes next. Celia explains why the quantum market is growing faster than even the companies inside it predicted, with revenues rising roughly 27% year over year and actual numbers consistently beating forecasts. She also makes clear that the future is not quantum replacing classical computers. It is hybrid systems combining both to solve problems that simply cannot be solved today, with early use cases already emerging in pharmaceuticals, energy, finance, and defense. We also get into quantum sensing, the most underrated corner of the quantum world. From biomedical imaging already in clinical trials to quantum clocks powering GPS and financial transaction timestamping, sensing is already partially commercialized and quietly reshaping industries most people have never connected to quantum at all. Finally, Celia addresses the AI question directly. Will AI replace quantum? No. The two are complementary. AI is already accelerating quantum hardware design and algorithm discovery, and quantum may eventually improve how AI systems are trained. She closes with a clear message for enterprise leaders: the transition to quantum will not be a migration. It will be a paradigm shift, and the time to start preparing is now. Subscribe for more conversations with the people building the future of AI and emerging technology. Stay Updated: Craig Smith on X: https://x.com/craigss Eye on A.I. on X: https://x.com/EyeOn_AI Timestamp: (00:00) Introduction: What Is QED-C and Why Does It Exist? (01:57) Celia Merzbacher on Her Background and Role (04:32) Annual Market Survey: How Fast Is Quantum Actually Growing? (09:10) Where Quantum Revenue Is Coming From Today (11:11) Timeline and the Race to Utility-Scale Quantum Computing (13:23) Early Use Cases: Pharma, Energy, Finance and Hybrid Computing (16:14) What Is Quantum Sensing and Why It Matters (20:39) The Three Pillars: Hardware, Error Correction and Algorithms (27:40) How Enterprises Should Start Preparing for Quantum Now (38:39) AI and Quantum: Allies Not Competitors
Transcribe →#340 Steffen Cruz: Training AI Without Data Centres
4w ago00:46:25Tap to summarizeWhat if you could train a frontier AI model without building a single data centre? In this episode of Eye on AI, Craig Smith sits down with Steffen Cruz, co-founder and CTO of Macrocosmos, to explore a radical alternative to the way AI models are built today. Instead of billion-dollar GPU warehouses, Steffen is training large language models using idle compute from devices distributed around the world, coordinated through the Bittensor blockchain. Steffen breaks down why the centralised data centre model is heading toward a wall. Projects like Stargate and Colossus cost tens of billions of dollars, and as appetite for larger models grows, the economics simply stop making sense. He explains how distributed training flips this on its head, tapping into surplus energy, underutilised GPUs, and even consumer devices like Mac Minis to train models at a fraction of the cost. We also get into IOTA, Macrocosmos's flagship technology, an orchestration layer that takes compute nodes scattered across the globe and makes them act like a single supercomputer. No single device runs the full model. Instead, each one carries a small slice, a technique called model parallelism, and together they can train frontier-scale models that would otherwise be out of reach for startups, researchers, and enterprises. Finally, Steffen shares what he's building toward: 70 billion parameter models trained at 10 to 20 percent of centralised costs, a two-sided marketplace for compute, and a future where anyone with a spare GPU or Mac Mini can earn passive income while contributing to the democratisation of AI. Subscribe for more conversations with the people building the future of AI and emerging technology. Stay Updated: Craig Smith on X: https://x.com/craigss Eye on A.I. on X: https://x.com/EyeOn_AI Timestamp: (00:00) Introduction: The Problem With Blockchain AI Projects (06:39) Meet Steffen Cruz: From Subatomic Physics to Decentralised AI (09:16) What Is a Bittensor? The Blockchain Built for AI (11:53) How the Blockchain Actually Works: Registry, Clock, and Rewards (15:08) Why Data Centres Are Hitting a Wall (22:01) Distributed Training vs Federated Learning: What's the Difference? (27:47) Train at Home: Turning Your Mac Mini Into a Passive Income Machine (32:49) IOTA Explained: Building a Global Supercomputer From Spare Parts (39:43) How the Network Scales: From 256 Nodes to Limitless Compute (44:39) The Road Ahead: 70B Parameter Models and the Future of Affordable A
Transcribe →#339 Eamonn Maguire: Your Child Has a Data Profile Before They're Born
4w ago00:45:30Tap to summarizeWhat if your child already has a data profile, and they haven't even been born yet? In this episode of Eye on AI, Craig Smith sits down with Eamonn Maguire, Director of Engineering for AI and ML at Proton, to explore one of the most urgent and underappreciated questions in the age of AI: who owns your data, who is building a profile on you, and what can actually be done about it? Eamonn brings a rare combination of depth and range to this conversation. With a PhD from Oxford, a postdoc at CERN, and years at Facebook engineering ML systems to detect internal and external threats, he now leads Proton's AI efforts, including Lumo, their end-to-end encrypted alternative to ChatGPT. He makes a compelling case that the surveillance economy is not just a privacy problem but a behavioral one, where the systems profiling you are not only observing who you are but actively shaping who you become. We get into how just three data points are enough for advertisers to infer your age, political leanings, religion, and spending habits. We discuss why trusting mainstream AI platforms with sensitive data is a structural problem, not just a policy one, and why the AI labs with the best models got there by acquiring the most data, often with little regard for copyright law. Eamonn also breaks down the difference between truly open models and open washing, and explains how Proton builds AI that is genuinely private by design, with local indexing, encrypted memory, and user-controlled data sharing. Then there is Born Private, Proton's initiative to give children a private digital identity from birth. It sounds simple on the surface, but the conversation it opens up is anything but. Data collection on your child begins before they are born, the moment a parent emails a gynecologist or a fertility clinic. Eamonn argues that until we start thinking about privacy the way we think about other rights, from the very beginning, the surveillance machine will always have a head start. Subscribe for more conversations with the people building the future of AI and emerging technology. Stay Updated: Craig Smith on X: https://x.com/craigss Eye on AI on X: https://x.com/EyeOn_AI Timestamp: (00:00) Introduction and Meet Eamonn Maguire (00:38) From Bioinformatics to CERN to Facebook: Eamonn's Career Arc (05:23) How Proton Started in the CERN Cafeteria (09:23) What Mainstream AI Platforms Actually Do With Your Data (13:00) Copyright, Training Data, and Why Big Labs Can't Be Trusted (15:10) Open Models vs Open Washing: What Truly Open AI Looks Like (24:22) How Lumo Works: Encrypted Memory and No Data Leakage (31:18) Born Private: Reserving a Private Email Address at Birth (33:00) How Data Profiling Starts Before Your Child Is Born (34:26) How Three Data Points Become a Complete Profile (39:07) Molly Russell and the Consequences of Algorithmic Profiling (53:55) The Full Proton Ecosystem: Mail, VPN, Drive, Lumo, and Workspace
Transcribe →#338 Amith Singhee: Can India Catch Up in AI? IBM's Amith Singhee on What It Will Take
Apr 2400:46:54Tap to summarizeWhat if the country that trains the world's engineers finally built the infrastructure to match its talent? In this episode of Eye on AI, Craig Smith sits down with Amith Singhee, Director of IBM Research India and CTO of IBM India and South Asia, to explore where India actually stands in the global AI race and what it will take to close the gap. Amith gives an honest, ground-level assessment of why India has been slow to compete. The talent has always been there. But until recently, the investment, the compute infrastructure, and the institutional intent hadn't come together in a sustained, coordinated way. That's changing, and Amith explains exactly what's different now. He walks through IBM Research India's 27-year presence in the country, the research it's doing on foundation models, hybrid cloud AI deployment, agentic systems, and quantum computing. He also explains why building AI from India doesn't just help India. Working with less data, less compute, and more linguistic diversity forces better engineering and makes IBM's models more generalizable for the entire world. We also get deep into the technical frontier. Why catastrophic forgetting is one of the key unsolved problems standing between current AI and anything more capable. How IBM is already shipping continual learning in practice through its COBOL modernization tools, helping enterprises decode decades of legacy code before the engineers who wrote it are gone. And why agentic AI, for all the hype, still has a mountain of unglamorous enterprise engineering left to climb before it becomes truly reliable. Plus, what Amith would tell an 18-year-old engineer in India today about what skills will actually matter in an AI-driven world. Subscribe for more conversations with the people shaping the future of AI and emerging technology. Stay Updated: Craig Smith on X: https://x.com/craigss Eye on A.I. on X: https://x.com/EyeOn_AI (00:00) Introduction and Amith Singhee's Background (06:26) Why IBM Set Up Research in India (11:45) Can India Compete in AI (15:18) How IBM Collaborates With Indian Universities (19:25) Why India Has Been Slow in AI (24:50) IBM's Hybrid Cloud AI Research Focus (27:34) How Data Scarcity in India Makes Better AI (31:18) Fine-Tuning Models Without Losing General Knowledge (35:03) Continual Learning and Catastrophic Forgetting (38:25) COBOL and Legacy Code Modernization (42:11) Agentic AI Hype vs Enterprise Reality (48:09) What Young Engineers Should Study Today
Transcribe →