Unscripted with Jeff Pedowitz

Matthew Guggemos spent more than 20 years as a speech-language pathologist teaching children to communicate — kids for whom language doesn't come free — and now builds AI at scale as CTO of iTherapy, the company behind InnerVoice, an AAC app with over 50,000 users. He's also a conservatory-trained drummer who played in Michigan State's top jazz band, and he joins Jeff to explain why music, language, and AI are all the same problem: getting nearly infinite expression out of a finite set of constraints.

Camberley Bates has spent 30-plus years as one of enterprise technology's most respected independent analysts, paid to cut through vendor hype and tell executives what's real. On this episode of Unscripted, she and Jeff unpack what IBM's earnings dip really signals (AI infrastructure spending delayed mainframe upgrades, not weakness), why 70 to 80 percent of enterprise data still lives on-prem, and why cloud-first has given way to workload placement. She explains the physics nobody can outrun: latency and data gravity make moon-based data centers a fantasy and on-prem AI stacks a boom for Dell and HPE. Bates gets candid about prompt privacy (she wouldn't bet enterprise data stays out of training models), the SaaS valuation crisis, the data center NIMBY wars, and how to decode "AI-ready" marketing. Her prediction for the claim that ages worst: that AI eliminates work. People need work to have value in their souls.

Russ Wilcox, CEO of ArtifexAI, went from a broken sewer decision in his Cape Cod hometown to advising the Pentagon and State Department on the US-China AI race. In this episode of Unscripted, he tells Jeff how mining boring public records, zoning codes, meeting minutes, and permit archives, unlocks the 85 percent of human knowledge trapped in unstructured data. He explains why he believes the US is currently losing the AI race: China treats AI as infrastructure, with data reservoirs and specialized swarm agents, while the West chases god models built on 15 percent of human knowledge. Wilcox maps China's chip packaging dominance, makes the case for cognitive sovereignty, and lays out his three-part playbook for bringing AI to local government: compliance, decision support, and transparency. His vision: AI that revives local democracy through public-private partnership instead of digital techno-feudalism.

Every impressive AI demo sits on an enormous, expensive, and very unglamorous foundation of data. Keith Pijanowski is one of the people who built it. After a decade-plus at Microsoft (where he helped evangelize Azure before its public release) and data pipeline work at BNY Mellon, he now works on the storage infrastructure that AI training and inference actually depend on. He breaks down why hundreds of GPUs hammering storage for months makes training a workload the software industry has never seen, and why inference is no longer lightweight thanks to KV caching across GPU memory, system memory, and network storage. He predicts power ratings will soon rival performance ratings for GPUs, dismisses space data centers as science fiction while taking ocean data centers seriously, and argues the build-out is rational because the Jevons paradox guarantees efficiency gains produce more demand, not less. His deepest concern isn't the machines. It's people publishing content and shipping code they don't understand. Use AI as a productivity tool, but don't let it turn you into an operator.
 

Everyone is promising humanoid robots in every home and factory. Klas Nilsson has spent 40 years on the part that turns out to be hard. The Cognibotics founder, who built one of the first digital robot motion control systems at ABB in the 1980s, explains why less than 1% of the world's work is done by robots, and why the answer isn't a smarter chatbot. He walks through why the robotic hand remains unsolved (no existing design, tendon-driven or motorized, is future-proof), why LLMs belong at the task-definition level and never in the servo loop, and why robots are still programmed by position when real work is about force. He predicts another mild AI winter before the community learns which lessons matter, puts general-purpose humanoids seven to ten years out, longer for the average home, and argues the endgame isn't robots replacing craftsmen. It's craftsmen teaching robots by hand and selling that skill. Machines that help humans, so humans only do human thing

Lena Smart has spent 30 years securing the systems we genuinely cannot afford to have fail — the power grid, financial markets, and as longtime CISO at MongoDB, one of the most widely used databases on the planet. She's defended more high-stakes infrastructure than most people will ever see. And she's here to tell you that AI agents scare her more than any of it.
Not because they're malicious. Because nobody knows who owns them when they break.
We get into the shadow AI problem CISOs are walking into blind, why the board is asking for AI on Monday and the CISO finds out on Friday afternoon, what Lena actually does when she wants to stress-test a company's chatbot (hint: she asks it for a lasagna recipe), and why she's now helping build the first real certification and insurance standards for AI agents through AIUC.
And before any of that — she grew up in the housing projects of Glasgow, left school at 16, taught herself to code from library books, found a modem in a dumpster, and joined hacker bulletin boards under a male name. She still won't tell you what the name was.
Lena Smart, Unscripted. Let's get into it.

Dr. Sarah Chardonnens is a professor at the University of Freiburg with a PhD in the science of learning, and in this conversation with Jeff Pedowitz she takes on the question everyone has an opinion about but few have studied directly: does AI make us smarter, or just make us feel smarter while a real skill quietly erodes underneath. Drawing on research showing LLM error rates ranging from 16 percent on basic questions to as high as 60 to 70 percent on complex tasks, despite consistently fluent and convincing output, Chardonnens argues the effect of AI on a learner depends entirely on their existing expertise and when in the learning process the tool is introduced.
 
She walks through her Synapse model, four phases of learning, sensory input, network adaptation, participation, and storage and embodiment, and explains why AI can genuinely amplify an expert's thinking while quietly flattening a novice's, particularly when it short-circuits the struggle that actually builds neural architecture.
 
Drawing on her own background as a concert musician and martial artist, she connects the discipline and productive struggle required in both to what's now missing in unsupervised AI use, and closes with a simple framework anyone can apply immediately: think about why you're using AI before you start, stay aware of whether you're still thinking during it, and ask afterward what you could still explain yourself, without the machine.

Juergen Weichenberger has been building AI since the 1990s, well before it became a mainstream conversation, in domains where the cost of being wrong isn't a typo, it's a shutdown: factories, energy grids, and critical national infrastructure.
 
Having led AI at Schneider Electric and now serving as a data and AI partner at EY, he tells Jeff Pedowitz why the industrial sector remains the least AI-penetrated part of the economy despite having some of the highest value at stake, tracing it back to a fundamental trust gap between engineers and AI teams.
 
The conversation moves through why unconstrained optimization can push a system to a dangerous edge, illustrated by a real example of AI-optimized throughput collapsing a company's own market margins, and why a widely repeated use case like predictive maintenance often adds zero value once you understand how much redundancy engineers already build into critical systems.
 
Weichenberger walks through a real project where 150 AI use cases recommended by a major consulting firm collapsed to a single viable one the moment they were validated against actual operations staff, and closes with what he considers the most underrated skill in AI right now: context engineering, and simply talking to the people doing the work before proposing a solution to a problem they may not actually have.

Manas Talukdar has spent 19 years building the infrastructure underneath AI systems: the data backbone of the process industry, the platform behind one of the largest enterprise AI companies, and the training data systems behind modern language models.
 
In this conversation with Jeff Pedowitz, he argues that the model, the part everyone argues about, is actually the easy part. The hard part is the data engineering, context management, and system architecture surrounding it, and that's where most enterprise AI initiatives quietly fail, particularly on long horizon problems where accuracy degrades as workflows get longer regardless of how large the context window gets. Talukdar walks through why reliable, production-grade agent fleets barely exist yet despite how good the demos look, why the industry's shift from token maxing to token optimization is really an economics problem in disguise, and why intellectual property exposure, not just cost, is quietly shaping how enterprises think about feeding their proprietary knowledge into third-party models.
 
He closes with a concrete picture of what an AI-native company actually looks like in practice, and lays out where he believes durable competitive advantage will live once the models themselves are fully commoditized: proprietary data, deeply integrated workflows, and operational discipline.

Ed Addison took his first AI course at MIT in 1985 and has spent the four decades since building through every wave of the field, from brittle rule-based expert systems in the 80s, to neural nets in the 90s, to the deep learning breakthrough of 2006, to the generative AI and transformer architectures that produced today's large language models.
 
He's founded five AI companies with three exits, most of it in the notoriously difficult world of AI drug discovery, where he argues the algorithms getting all the attention aren't actually the valuable asset: the data is, and Big Pharma still owns most of it. In this conversation with Jeff Pedowitz, Addison lays out why no AI-discovered drug has made it through FDA approval yet, why failure rate rather than speed or cost is the only number that will prove the technology works, and why he believes the AI industry has a financial bubble but not a scientific one.
 
The conversation closes with the story behind his "accidental novel," Probability of Doom, written with AI as a collaborator after a family road trip conversation, and now used in his own classroom to teach engineers what can go wrong when humans, humanoids, and multi-agent systems start operating side by side.