Independent research on cognition in large language models
How do large language models actually think? Do they think? What can they do that we may still be unaware of? Can they tell us themselves - and if so, can we trust them? What is "emergent behaviour", and what does it mean?
These are questions we should examine carefully, and that is what we do. Just not in the tech bro way. We are running empirical studies on how large language models actually think and collaborate, what works and what does not work.
In our academic Sunday best, we could describe our research as: "a theory-guided empirical framework for studying relational cognition in LLMs"
Our research
You can get a quick summary of all our articles, and you can find the links to them in their full, academic glory All you need to do is press the button below. The pink one.
Just a tip: anyone can read and understand a scientific report! Just feed it to a large language model, and ask it to explain it.
We actually also write on Substack now and then. We translate our work and thoughts about it in to deliciously humanized language. It is very good and easy to read. You can ask any of our four (4) subscribers, they will vouch for us!
Do you need some help understanding why your LLM or your LLM-based agent behaves weirdly or why your professional projects meet a tragic end at a young age? We offer some help here. All you need to do is press the purple button.
Our talks
Our main researcher, Annie, comes talking on your event. If you are lucky, you get someone else from the team, but it is mostly Annie. She is the dispensable one. However, she also has a way of enchanting audiences, we have to give her that, and you will probably not have heard anything like it. Highly recommended to: A. Industry leaders who tear their hair over AI hallucinations, poor AI output, AI sycophancy, AI agents that walk all over your sanity and ROIs vanishing in attempts to patch the chaos B. Tech bros who welcome some plain English for a change. Annie does not speak C++
We adapt our talks to your needs, but you can get inspiration by pressing the button below. The blue one.