Talks
A list of ready-made talks I have given on multiple occasions. I am usually happy to develop more tailored content and workshop formats on topics within my research areas. I offer talks in English or Danish. I like to think I can communicate any of my research to pretty much any audience, so feel free to contact me if you are interested in booking me.

The Cognitive Cost of AI
Danish: De Kognitive Omkostninger ved AI.
What happens to our cognitive functions when we use generative AI to outsource knowledge work?
This talk takes a skeptical look at generative AI and what it means for human cognitive abilities and capabilities to use generative AI in knowledge work - work that depends on essential human skills such as creativity, knowledge, and analytical thinking. I present current research in the area as well as hypotheses for the future of human cognition in the age of AI.
​

AI "Agents" - Security and Safety
Danish: Sikkerhed i AI-agenter.
A talk about the difference between security and safety, and how we might approach either in the context of AI "agents"
The fundamental principle of the AI "agents" paradigm is that programs need not be limited to being controlled by humans to complete tasks, but that we can create programs that can "autonomously" combine and perform functions without human intervention. The problem is; errors cascade and risks amplify with the level of autonomy. Risks to safety and security is the price these tools charge for the productivity they deliver!​

Hacking for AI Literacy
Danish: Hvad hacking kan lære os om AI Literacy.
A workshop with departure in the arcade machine Hacc-Man - a game for teaching AI literacy and LLM self-efficacy through hacking
What if, instead of teaching about the virtues of ChatGPT, we taught about its weaknesses? The Hacc-Man arcade machine is a game for anyone amd involves hacking your opponents: six different chatbots. Each chatbot comes with a different challenge - for example, coaxing out the mayor's home address or persuading a narrator bot to output NSFW content. This talk can be more or less interactive and last anything from 45 minutes to 4 hours.
Note: if you want me to bring the arcade cabinet, you must cover the transport cost. I have a great transportation supplier though, it's not as expensive as I would have thought! If not, I can bring a local version of the game.
​​

LLM Security - How People Attack Large Language Models
Danish: LLM-sikkerhed - Hvordan man hacker sprogmodeller.
A talk based on our PLOS One paper: Summon a Demon and Bind It: A Grounded Theory of LLM Red Teaming
In this talk I report the findings of a large qualitative study of formal and informal LLM red teamers. This includes looking into the motivations for red teaming: why do people do it, and what do they learn? I also go through 12 strategies and 35 LLM red teaming techniques. As a bonus, I offer an unhinged take on why jailbreaking and prompt injection are the wrong terms for these practices. And I discuss why metaphors of magic and demon summoning are useful for talking about these activities.
​
Image credit: Promptmancer by Feddie Xeth.​​

Under the Hood of "AI" - Stochastic Parrots and Mechanical Turks
Danish: I have yet find a good translation for stochastic parrots or mechanical Turks!
A critical perspective on the metaphors we use to describe so-called artificial "intelligence"
Language matters. How we describe "AI" technology influences how it is perceived, deployed, and trusted. In this talk, I present the argument that using anthropomorphizing metaphors to describe these technologies is misleading and incorrect, and I give examples of other metaphors that profoundly change how we perceive and act in the world. I suggest strategies for how to spot and combat the "AI" hype by focusing on describing what the technologies help us achieve, rather than what capabilities we posit they have.​​

The Tools We Think with and What they Teach Us
Danish: Værktøjer Vi Tænker med og Hvad De Lærer Os.
An overview of my research in cognitive augmentation tools
In this talk, I go over my main research strands:
1. Tools we use to teach: Tools that are purposefully used to teach subject matter, for instance coding tutorials and online or e-learning.
2. Tools we use to learn: Digital tools that are used ad hoc to process and manage knowledge, for instance in note-taking, reading, and information management.
3. How tools teach. I look at what tools teach users about themselves by mode of their design or their description.
​​

Where Do Ideas Come from?
Danish: Hvor kommer idéer fra?
A talk about the science of creativity
This talk is about how ideas are formed in the brain and what characterizes creative thinking. With a departure in fundamental creativity research I suggest different strategies for how we best facilitate the creation of truly creative ideas and why this approach makes sense based on neuroscience.
​​