Real self's and meaningful language outside biology
potential biological minds ≈ potential machine minds, trying to remove 'potential'
Jonas Strabel
Robotics, Cognition and Intelligence at TUM while working in Data Analytics / Data Science at MedifoxDan (ResMed). Before that: applied computer science at DHBW Mannheim alongside work at Roche Diagnostics and philosophy at PTH Sankt Georgen.
More informationDownload CV (PDF) ↓
Writing
All writing →Can large language models know what they know?
Sosa's account of aptness meets mechanistic evidence and the second-order problem of whether an LLM can recognize its own competence.
Is unregulated AI research a good idea?
The familiar pause-or-accelerate argument changes depending on whether AI development is an arms race, a prisoner's dilemma, or a tragedy of the commons.
Sometimes, LLMs think
A reliable-faculties route to machine thought, and a role for mechanistic interpretability in separating competence from coincidence.
Projects
All projects →Natural language acquisition POC ↗
Can an embodied learner acquire a self-model and make that model useful for both regulation and communication—such that even deflationalists will concede that it uses language meaningfully? ;D
AccountyCat ↗
A macOS focus app that looks at what I am actually doing before deciding whether I am procrastinating. Text first; vision only when the window refuses to explain itself; local models if preferred.
Web of Belief ↗
Eighteen questions, thirty-five philosophical positions, and a rule engine that points out where the answers clash. It judges consistency, not the person answering.
Robots101 ↗
ROS 2 and Gazebo running natively on macOS, with a TurtleBot that can navigate, see, and take structured plans from language and vision models. Mostly an exercise in making incompatible layers admit they know each other.