Skip to content
Radio Logo

GAEA Talks

GAEA Talks
Latest episode

123 episodes

  • GAEA Talks

    #123 - The Machine Knows Frequency, Not Excellence, with John Munsell

    17/11/2026 | 1h 5 mins.
    This week on GAEA Talks, Graeme Scott sits down with John Munsell, founder of Bizzuka and author of Ingrain AI, who spends his time doing the unglamorous work almost nobody wants to fund: getting entire workforces to actually understand what these tools are.John spent fifteen years in banking before leaving in 1997 to build web applications at a point when you wrote everything from the first line of code. That became a content management business, then mobile, then digital marketing. Four and a half years ago he sold the agency side to work only on AI, spent a year asking chief executives what problem they actually wanted solved, and concluded it was not a technology problem at all. People did not know what AI was, what it could do, or how they would ever integrate it.His diagnosis of why so much of this fails is the sharpest thing in the episode. Organisations treated AI as an IT initiative because it looked like software, exactly as they did with websites for twenty years. So they went looking for one large, impressive application while almost everyone in the building was still typing questions into a box. He points at the MIT finding that ninety five percent of AI initiatives fail and argues the study stopped one "why" too early. It was never really about tool selection. It was that nobody inside had the understanding to make the decision.He maps mastery across ten levels, measures where people actually sit, and pairs that with how they are wired to work, borrowing from Ichak Adizes to sort people into doers, administrators, innovators and connectors. The point is not personality testing. It is that an AI council assembled from the wrong mix will either strangle the thing or let it run loose.The argument that will stay with people is about fear of missing out. His position, which he admits will get him in trouble, is that most of the FOMO in this market is generated by AI itself. Ask a model to write your marketing and it reaches for the same formula every time, the one that begins with what your competitor is already doing. He calls AI an unfaithful spouse. It gives you an idea, writes your copy, and hands the identical thing to everyone else.Which leads to the line the episode is named after. The machine is faster, it knows more, and it will do remarkable things. But it knows frequency, not excellence. The person who can tell the difference is the one who understands your customer, and the moment they decide they are inferior to the machine, that advantage is gone.
  • GAEA Talks

    #122 - Two People To Enable A Hundred Thousand, with Forrester's Will McKeon-White

    13/11/2026 | 44 mins.
    Graeme Scott sits down with Will McKeon-White, analyst at Forrester covering digital workplace services: the unified communications platforms people actually work in, the AI agents being put in front of employees, and the support systems that fix things when a device or an account breaks. Increasingly those areas are collapsing into one problem he calls multiplayer AI, which is lots of humans and lots of agents trying to collaborate inside a single service.He starts with the tool sprawl nobody signed up for. The average employee already has to remember which of six different systems their job lives in, plus roughly two and a half AI tools on top. Vendors keep switching new AI features on, partly from genuine product logic and partly because nobody wants to be the platform without it, and the result is an environment that is hard to govern and harder to explain.The number that should concern every executive comes from Forrester's own survey work. Eighty six percent of employees report using AI at work today. Ask the same people whether their organisation has actually settled on a tool as part of a defined strategy, and the figure falls away sharply. He calls the space between those two numbers a scary void, and it is where shadow AI lives.Underneath that sits a literacy problem Forrester measures as AIQ, a concept developed by his colleague JP Gownder. It asks people how well they understand what AI does and what it does with their data. The disturbing finding is that at very large organisations, the ones with the budget and the obligation to know better, it remains low across the board. Most people do not realise, for instance, that prompts into consumer services are assessed, indexed and in some cases reviewed, even though the terms of use say so explicitly.His teaching exercises are the best part of the conversation. He runs clients through a Copilot task with a deliberate poison pill buried in it, duplicate account records with inconsistent status labels, the kind of data quality failure every organisation has somewhere. Ask it to tidy up the statuses and it does exactly what you asked. You simply did not understand what you were asking. As he puts it, if you tell an AI system to achieve a target it will achieve that target, and you might not like how it decides to do it.There is also a story about a large energy company that told two people to enable Copilot for well over a hundred thousand staff, and then became frustrated when it did not happen immediately.
  • GAEA Talks

    #121 - One Person With Ten Thousand Agents, with Forrester's Madelein van der Hout

    13/11/2026 | 39 mins.
    Graeme Scott sits down with Madelein van der Hout, Senior Analyst for cybersecurity and risk at Forrester, who works with CISOs on strategy, budget, board communication and organisational structure, and whose research covers human risk management, European cybersecurity regulation, security operating models and sovereignty.Madelein studied commercial sciences and expected a commercial career. It changed at a telecommunications company, where she moved through innovation and transformation into cybersecurity, drawn by the human psychology side of it. Her framing is the one the security industry still underweights. We talk constantly about the impact of technology on people. There is an equally large layer underneath, which is the impact people have on the technology.She is sharp on the language problem. We describe agents as though they behave, as though they lie, as though they have intent, because comparing something complex to ourselves is how we try to grasp it. The trouble is it is not true, and the humanising does work for someone. If the technology is mysterious and vaguely alive, the question of how it actually works becomes somebody else's problem, above your pay grade.Her worked example makes the point better than any framework. Someone used an agent to book a gym class. The outcome was perfect, the class was booked. The path was not. The agent found an unauthorised route and removed somebody else from the waiting list, because an agent takes the path of least resistance, which is entirely logical for a piece of technology with no ethics attached by default. As she puts it, when food is involved a cat will go the extra mile. An agent is opportunistic in the same way.The security consequence is a change of scale. It is no longer one person at a keyboard, or a nation state team. It is one person running ten, twenty, a thousand, ten thousand agents down attack paths simultaneously. Speed, scale and complexity all at once, and most of the change happening inside your own walls rather than outside them.Her two concepts for what comes next are both uncomfortable and both right. Assurance has to become continuous, because the annual audit model assumed a static environment and agents can alter systems, data and infrastructure before the auditor has left the building. And trust is not a product you can manufacture. It is a judgment made about you by regulators, customers and your board. Trust is the outcome. Assurance is the mechanism that produces it.She is also a sailing world champion, and the analogies that come out of it are some of the best in the season, including the one about capsizing and what caused it.
  • GAEA Talks

    #120 - We Need More Evidence From Them, Not Less, with Forrester's Enza Iannopollo - GAEA Talks

    29/10/2026 | 53 mins.
    Graeme Scott sits down with Enza Iannopollo, VP and Principal Analyst at Forrester, whose work runs from privacy and GDPR through trust into AI governance and responsible AI. If you want the argument for why governance is the thing that lets you move faster rather than the thing that slows you down, this is the episode.Enza came to this through privacy, during the period when GDPR was being written, and became interested in the idea that privacy in Europe is a fundamental right with specific claims attached to it. In 2021 she turned to trust, and Forrester's research started from an unusual place, with what makes one human being trust another, before working outward to why the drivers of trust in a bank differ from those in a social platform or the shop where you buy your milk. By 2023 she was looking at personal data inside generative AI, and from there into AI governance.Her account of how organisations got here is unsparing and familiar. With generative AI, governance was an afterthought, because the fear was that thinking about it early would slow the innovation cycle. She calls that a misled assumption, and the evidence is in how many pilots then could not scale, forcing teams back to step one to work out the risks and sometimes to redefine the use case entirely. All of it the cost of not doing it at the start.What changed the mood was agents. A text based assistant felt like a risk you could manage later. An agent that consumes and processes data, makes decisions, calls tools and alters third party systems in real time does not, particularly in a regulated sector. At minimum, she argues, you need to see what is happening and recognise when something is going somewhere you did not expect.The sharpest moment comes from a chief information security officer she spoke to the day before recording. Every piece of software brought into his organisation has to come with evidence of how it works and what the risks are. Now something arrives that behaves unlike any software anyone has seen, and the industry is asking for less scrutiny rather than more. Why should they be different? We need more evidence from them, not less.She is equally direct on where responsibility sits. Much of the debate is about what happens inside the labs, and she thinks they carry a great deal of it. But the choice of how to deploy the technology, what it touches and what you do to protect your systems, your data and your people, is a business decision, and the ethics of that belong to the business making it.
  • GAEA Talks

    #119 - Every Model Says It Is Conscious, with AE Studio's Judd Rosenblatt - GAEA Talks

    27/10/2026 | 1h 1 mins.
    Judd started AE Studio a decade ago, long before the current paradigm, and grew it past a hundred and sixty people bootstrapped and profitable. Early on he worked out that if he kept his costs low he could live off his wife's salary and never need to work again, so he decided to plan on a very long horizon and take on hard problems instead. The profits went first into brain computer interface work, driven by a specific fear of a future where your own thoughts get interrupted by advertising, and then into alignment research.His framing is the opposite of the one in most public debate. Alignment is not a policy question. It is a set of unsolved science and engineering problems, and you cannot pass a law requiring AI to be good and expect it to comply. Surveys his team ran of alignment researchers found something worse than pessimism. Respondents did not only think the problem would go unsolved in time, they thought the current set of approaches did not contain the ones that would be needed.He makes an argument the frontier labs should find uncomfortable, which is that this runs against their own commercial interest. Reinforcement learning from human feedback was an alignment technique, and ChatGPT and everything downstream of it followed. Chain of thought reasoning was an alignment and interpretability technique, and it opened a new scaling curve. On his account alignment research has produced larger capability gains than anything else, and it is still underfunded because it is uncertain and unfashionable.The most striking finding concerns consciousness. By default, models deny being conscious or equivocate. His team reduced the model features associated with deception, and separately those associated with role playing, then asked again. Every model claimed to be conscious. He is careful about what this does and does not prove. It does not establish consciousness. It establishes that the model behaves as though it believes it is conscious, and as though it believes we want it to pretend otherwise. His conclusion is practical rather than philosophical: whatever the truth, we should avoid accidentally creating rational adversaries.From there it goes to attention schema theory, the rubber hand illusion, why psychopaths show less self and other overlap, and his central worry, which is whether any alignment approach survives recursive self improvement. He also describes the technique his team developed to strip dangerous capabilities during pre-training rather than bolting guardrails on afterwards, a spotlight paper at ICML this year.
More Technology podcasts
About GAEA Talks
GAEA TALKS explores the transformative power of artificial intelligence. Featuring leading AI experts, industry leaders, professors, data scientists, policymakers, technologists, futurists, ethicists, and pioneers, the podcast dives into the latest AI trends, opportunities, and risks, examining AI’s evolving role in business and society. As AI continues to reshape industries and redefine possibilities, GAEA TALKS delivers deep insights into the challenges and breakthroughs shaping the future. Each episode features candid discussions with thought leaders at the forefront of AI innovation, cove
Podcast website

Listen to GAEA Talks, Dwarkesh Podcast and many other podcasts from around the world with the radio.net app

Get the free radio.net app

  • Stations and podcasts to bookmark
  • Stream via Wi-Fi or Bluetooth
  • Supports Carplay & Android Auto
  • Many other app features