Artificial intelligence is no longer a topic of the distant future. It is today an integral part of our lives and has the potential to transform society and technology profoundly. That is exactly where we come in with our podcast Attention Heads. For over a year, we, two passionate AI developers, have been bringing engaging insights, well-grounded analysis and stimulating discussions to you every two weeks.
Attention
Heads
Podcast: Swiss perspectives on AI and the technology of tomorrow
Attention Heads is hosted by my long-standing friend and colleague Joel Barmettler and me. Joel and I have known each other since our student days and have in the past co-founded a startup together, which taught us important entrepreneurial and technical lessons. Today we work together at a different company, where we continue to deepen our passion for technology and artificial intelligence. But our relationship goes far beyond professional collaboration: Joel is not only a brilliant thinker but also a close confidant with whom I have been exchanging ideas and realising joint projects for years.
Our goal is to illuminate the fascinating and often misunderstood world of AI from a unique Swiss perspective. While many public debates are shaped by fears and concerns, whether through dark visions of the future or supposedly uncontrollable dangers, we want to put these discussions on a factual and technical footing. What does AI really mean for our society, especially for Swiss democracy? Where lie the opportunities, and where the actual risks? And above all: what is relevant today, and which developments might only play a role in 10 or 20 years?
In each episode we address the central questions around AI, from ethical challenges through groundbreaking technological innovations to the question of how AI influences our daily lives and our political systems. We separate science fiction from reality, put current topics in the right temporal context and create a space for honest, well-grounded discussion.
ChatGPT Would Vote for the SP
The Political Orientation of AI
Large language models like ChatGPT, Gemini, or Claude are no longer just technical toys: they influence how we consume information, conduct political debates, and make decisions. But how politically neutral are these models really? Do they carry an inherent ideological colouring? And if so, which?
80min min.
DeepSeek R1: Independent Thinking in Action
How AI Shapes Its Own Thinking Process
In this episode we take a close look at the groundbreaking DeepSeek R1: a model that independently defines and optimises its own thinking process. Instead of being told how to think, we observe how the AI generates its own chains of reasoning in real time and solves complex problems along the way. We explain how DeepSeek R1 uses reinforcement learning and innovative methods to pave the way to autonomous reasoning, thereby challenging traditional models that depend on human guidance. Join us as silent observers experiencing the fascinating self-organisation and free thought process of this open-source AI, and discuss what this means for the future of intelligent systems.
65min min.
SRF Arena: Does Radical Expropriation Lead to Digital Democracy?
Part 1 [1/3]
In the SRF Arena the controversial proposal to expropriate large technology companies was among the topics discussed: a radical approach aimed at a fundamental redistribution of digital power. In this reAct episode we examine exactly this point. We analyse whether such an intervention could pave the way to greater digital democracy, what opportunities it opens up, and what risks are involved. Join us as we critically question the arguments put forward in the Arena and place them in a technical context.
78min min.
SRF Arena: Less Work for Everyone - or More Wealth for a Few?
Part 2 [2/3]
This episode is dedicated to the far-reaching changes in the world of work brought about by the use of artificial intelligence. We pose the provocative question: does automation lead to more leisure time and a higher quality of life for the broad population, or does ultimately only a small number benefit from a digitally generated surplus? With a perspective that lightly incorporates Marxist approaches, we analyse how jobs and wealth distribution could really develop in the era of AI.
106min min.
SRF Arena: A Global Safety Mechanism for AI?
Part 3 [3/3]
In the SRF Arena the question was also raised of how the rapid progress of artificial intelligence can be regulated safely and responsibly. Experts discussed whether national measures are sufficient or whether a global safety mechanism, comparable to approaches in nuclear technology, would be necessary to minimise risks. In this reAct episode we engage with the proposals put forward in the Arena and ask: can an internationally coordinated regulatory framework really secure the digital future? We examine the opportunities and challenges of such an approach and place it in a broader technical context.
74min min.
o3: Does AI Surpass Human Intelligence?
Part 1 [1/2]
In the OpenAI video with Sam Altman the new model o3 was presented, and it is causing a stir on the Arc AGI benchmark. In this reAct episode we watch the video and react live to the impressive demos. We discuss whether o3 really approaches or even surpasses human intelligence on complex mathematical and programming tasks. Join us as we scrutinise the performance figures and challenges shown and assess whether a genuine quantum leap in AI development is visible here!
48min min.
o3-mini: How o1 Becomes Obsolete in Record Time!
Part 2 [2/2]
The OpenAI video with Sam Altman presents not only the powerful model o3 but also the new o3-mini, a model that scores with an ingenious cost-to-intelligence ratio. In this reAct episode we watch the video and react to the spectacular benchmarks and demos that show how o3-mini delivers such outstanding performance despite higher costs that o1 becomes practically obsolete. Join us as we watch the new standard in AI performance being set and discuss what this means for the rapidly evolving AI scene!
42min min.
When AI Kills
Dangers That AI Brings With It
Which dangers brought by AI are already reality today, in 2025? We report on where we really need to keep a close eye on AI developments, and use examples such as "Predictive Policing" and "AI in warfare" to show why the use of AI already poses genuine dangers today.
71min min.
tldr; Are We Being Manipulated by ChatGPT o1
Days 1 to 5 of 12 Days of OpenAI
We talk about 12 Days of Shipmas! More specifically about the new model o1 and ChatGPT Pro, about reinforcement fine-tuning, Sora, Canvas, and Apple Intelligence.
83min min.
tldr; At Google, AI Programmes Itself
A Remarkable Glimpse into the AI World at Google
A quarter of all new code at Google is now written by artificial intelligence: a remarkable glimpse into the rapid development of the technology. What does this mean for the future of programming and the role of developers? In this tldr episode we discuss Google's latest earnings report and the implications of AI already playing a significant part in creating software. A fascinating look at self-programming AI and what it means for the industry and the world of work.
46min min.
Why AI Projects Fail
and How You Join the Successful Ones
Poor planning, unrealistic expectations, and a lack of know-how: there are many reasons why AI projects stall or fail. In this episode we reveal what it really takes to make AI projects succeed. From the first steps of project planning through to implementation: find out which pitfalls to avoid and which strategies will help you join the winners. A must for anyone who wants not only to start AI projects but to see them through to success.
112min min.
The Secret Language of AI
A Look Behind the Scenes of Deep Learning
What really happens behind the scenes of Deep Learning? In this episode we lift the veil on the "language" of artificial intelligence and dive deep into the world of Embedding Space. We discuss how Dropout Layers improve AI learning, why Few-Shot Learning is so revolutionary, and how Fine-Tuning makes models more precise. But that is not all: we also take a critical look at Data Privacy and ask how safe our data really is in the world of machine learning.
66min min.
tldr; Why AI Must See in Order to Understand
Vision and Illusion
In this episode we show why artificial intelligence must see in order to truly understand. Vision models are the key to the next level of AI development, because only with images can a comprehensive picture of the world be built: image data surpasses text in quantity and depth. But that is not all: we also talk about the "Diffusion Illusion", a cool technique in which Stable Diffusion models can depict multiple perspectives and subjects simultaneously in a single image.
33min min.
abc; Artificial General Intelligence - in 1000 Days?
Offline on the Smartphone Thanks to BitNets?
In this episode of Attention Heads we dive deep into the world of artificial intelligence and ask ourselves: will it really be possible to have AGI (Artificial General Intelligence) in 1000 days? This bold statement comes from none other than Sam Altman. But what really lies behind it? We guide you from A to C through exciting concepts such as Attention Heads, Agents, and AGI, through technical milestones like Backpropagation and on to Contextual Embeddings. An episode that examines the future of AI and its fundamental building blocks.
56min min.
tldr; OpenAI o1:
The Next Step Towards AGI?
In this episode of tldr we take a closer look at OpenAI's newest model: o1. What makes this model different? With an alleged IQ of 120 and the ability to think before answering, it could be the next step towards AGI (Artificial General Intelligence). But how intelligent is it really? We discuss the most important innovations, how to get the most out of this model, and what challenges still need to be overcome on the path to genuine artificial intelligence.
37min min.
tldr; Artificial Self-Reflection
Strawberry and Reflection Get AI Thinking
While the AI industry is dominated by hype and marketing promises, we take a critical look at two current developments: OpenAI's mysterious "Strawberry" model and the open-source "Reflection" model. We question the big claims and analyse what really lies behind these new approaches.
51min min.
AI Dystopia 2035:
When AI Becomes the Lifeblood of the Economy (Double Episode: Part 1)
This episode critically examines possible negative effects of advanced AI over the next 10 to 15 years. We discuss scenarios such as massive job losses, changes in interpersonal communication, and potential concentrations of power among AI companies. Part 1 of our double episode on the future of AI: realistic threat or exaggerated fears? Part 2 will explore utopian visions.
47min min.
AI Utopia 2035:
When Algorithms Inspire Us to Greater Humanity (Double Episode: Part 2)
In this optimistic companion episode to our bleak AI dystopia we take a hopeful look at a future in which AI realises its full potential. From the four-day work week to a renaissance of community spirit, we explore how AI could enrich our lives. But be warned: this utopia will not come on its own. Find out what steps we need to take today to live tomorrow in a world where AI serves not only the economy but above all the people. An episode that will open your eyes to the positive possibilities of AI!
51min min.
AI Hype:
When Marketing Promises Meet Hard Reality
AI is on everyone's lips, but what really lies behind it? We critically analyse the current AI hype, question inflated marketing promises, and discuss the gap between expectations and reality. Along the way we examine the role of AI in companies, society's expectations of AI systems, and the driving forces behind AI adoption. A sober assessment of the AI boom with a look at future developments.
47min min.
Open Source AI:
The Forgotten Foundation of the AI Revolution
This episode looks at the often overlooked importance of open-source AI models. We discuss how these models drive innovation and research and why they are essential for the future of AI. Topics include OpenAI's move away from an open-source approach, challenges and opportunities in open-source development, and a comparison with proprietary models like GPT-4. The episode highlights the variety of available open-source models and their potential for democratising AI technology. For anyone seeking a deeper understanding of the current AI landscape.
47min min.
Is AI Intelligent?
An Answer Nobody Needs
In this episode of "Attention Heads" we tackle the much-discussed but perhaps unnecessary question: is artificial intelligence really intelligent? We philosophise about the definition of intelligence and analyse whether AI meets those criteria. In doing so we arrive at the conclusion that the question itself often creates more confusion than clarity. We examine why it may be irrelevant whether AI can be described as "intelligent" and shift the focus to the practical applications and ethical implications of the technology. With a touch of self-irony we question whether our obsession with the intelligence of machines is not more of a distraction from more important questions. Join us on this intellectual journey and decide for yourself: is the search for an answer to this question really necessary, or is it ultimately an answer that nobody needs?
55min min.
AI in the Classroom:
Revolution or Risk?
In this exciting episode of "Attention Heads" we tackle one of the hottest topics in modern education: artificial intelligence in the classroom. From universities to primary schools, AI is increasingly permeating the education sector. But what does this mean for students, teachers, and the education system as a whole? We discuss the use of AI like ChatGPT in teaching and analyse the potential benefits and risks. We examine the possibility of cheating through AI-assisted tools and ask whether they create an unfair competitive advantage. We also look at current assessment criteria and legal prohibitions around the use of AI in schools and universities. Find out how educational institutions can make sensible use of AI and which future scenarios are conceivable in which GPT models become a permanent part of learning. Join us in the debate and discover whether AI represents a revolution or a risk for education.
42min min.
Dangerous Biases:
The Invisible Influence of OpenAI on ChatGPT
This episode of our AI podcast ventures into a controversial and often overlooked topic: the inherent biases in AI models like ChatGPT, with a special focus on their effects on minorities. We reveal the profound yet largely unknown influence that OpenAI exerts on the development and design of ChatGPT. This episode illuminates how biases in the data used to train these AI models flow not only into the technology itself but can also affect interactions with users and society as a whole. Through conversations with experts and critical analyses we challenge our listeners to reflect on the ethical implications of AI development and discuss what can be done to shape a fairer and more inclusive AI future.
40min min.
DeepSeek V3: Open-Source Powerhouse
The Secret Behind the Hype
In this episode we take a close look at the DeepSeek V3 paper: the open-source powerhouse that is currently causing a stir in the AI scene. We explain why this model, with its 671 billion parameters (37 billion active per token) and innovative architectures such as Multi-Head Latent Attention and Mixture-of-Experts, is revolutionising the market. We examine how DeepSeek V3 achieves impressive results despite low training costs (only around 5.58 million USD) through efficient use of Nvidia H800 chips and lean data usage. Find out what lies behind the hype, what technical innovations distinguish the model, and why it is considered a game-changer in open-source AI.
61min min.
AI as a Job Killer?
How ChatGPT's Autonomous Agents Threaten the World of Work
In this episode of our AI podcast we confront a provocative question: will ChatGPT and its autonomous agents become a job killer for billions of positions? This episode dives deep into the discussion about the potential risks and far-reaching effects that AI-driven technologies could have on the global labour market. We examine how the advancing development and implementation of autonomous agents could threaten not only simple, repetitive tasks but increasingly also more complex and creative fields. Through a critical engagement with current trends, statistical forecasts, and expert opinions we explore how societies can prepare for a future in which work is redefined. The aim is to create an awareness of the urgency with which we must confront the challenges of AI integration into the world of work.
46min min.
AI Terms: B******t Check:
What Really Lies Behind Them
In this episode of our Swiss AI podcast we set ourselves the task of shedding light on the often confusing world of AI terminology. We untangle and explain the basics and key terms such as Artificial Intelligence (AI), Machine Learning, Data Science, and language models. This episode offers a clear and in-depth look at the meaning of these terms, their differences, and how they are connected. We also discuss how these technologies are developed and in which areas they are applied. The aim is to give our listeners a well-grounded understanding of these revolutionary technologies that goes beyond buzzwords. Whether you are new to the world of AI or an experienced developer, this episode promises to be informative and illuminating.
43min min.
ChatGPT: Curse or Blessing for Democracy?
Attention Heads - Swiss Perspective on AI and the Technology of Tomorrow
In this episode of our Swiss AI podcast we tackle a charged topic: the influence of ChatGPT on democracy. This episode dives deep into the complex interplay between artificial intelligence and democratic discourse. We discuss both the potential dangers and the possible benefits that ChatGPT and similar technologies can bring to democratic processes. Along the way we examine topics such as opinion manipulation, information dissemination, and the role of AI in political decision-making. Our goal is to develop a balanced understanding of how these innovative technologies can both challenge and enrich our democracy.
53min min.