Colloqium #45
Sep 19, 2026 at 12:00 pm CEST
Description
Last time, we talked about the difference between ‘alive things’ that cannot answer and ‘dead things’ that do answer.
This time we are asking ourselves, what can different theories of human consciousness tell us about the possibility of machine consciousness? And what can we learn from the hugging face incidents?
For the input and nudges to start up your thought processes about the topic, you can read up on the last discussion intro and check the material links at the bottom of the page –> Link zum 40th Colloquium: https://phencoco.net/archive/#40_aiandconsciousness
To sort the discussion a little, we have four sub-questions:
- Are machine consciousness and human consciousness comparable? We talked about these things before, you can check the Colloquium and material links for this topic here: Link to 7th Colloquium https://phencoco.net/archive/#artificial_ignorance_ux_digital_counter_enlightenment
- What is (human) consciousness, and is it digitisable, considering the inner phenomenological perspective and personhood? Link to 32th Colloquium: https://phencoco.net/archive/#ai_dreamnottechnology
- What is the difference between LLM´s and autonomous agents based on LLM´s? Link to 13th Colloquium: https://phencoco.net/archive/#aineuralnetworks
- Is the theory of the dividual a good explanation for machine consciousness? Link to 42th Colloquium: https://phencoco.net/colloquium/againstcognition/
A practical way to approach the topic might be the analogy of an autopilot´s attention and perception, leading to decision making in ‘autonomous cars’ and the conscious awareness and attention needed by people in a drivers seat. It could be useful to look at, because humanity has been experimenting with the automation of driving tasks for some time. Those experiments have re-shown the importance of attention, perception and decision making in a very high stakes and real environment. Those skills are clearly an important part of what we, even if untrained in matters of human cognition or autonomous cars, understand as consciousness.
Consider automated driver assistant systems; they are supposed to (and do, via face and eye tracking) remind you to watch the streets. If you are driving with the autopilot on, you are still the responsible party in the car - not the car ‘itself’ - and are supposed to be able to take over in high risk situations, as you as person have a better perception of any given complex traffic situation. Consider the back and forth of attention, which is happening here, and which is supposed to be both taken over, assessed and let go of, by one of the “attentive” positions/systems/agents. We did more than one talk on that, too: Link to 14th Colloquium https://phencoco.net/colloquium/aicognitiveassistance/
Additionally, there is a very good summary on the results of the technological assessment of the happenings around hugging face, lately: AI agents seemingly broke out of their non-internet-connected environments, and hacked the hugging face platform to compare, share and improve AI models. The incidents were debated as marketing stunts or AI with consciousness. Open AI published a 32 page report on the matter. At the same time, METR published a hugging-face-incident-report last month. You can watch the summary on you tube. Please find the Link to the video, the METR report and other materials below.
Lastly, there is a text by Luc Steels that at least goes through many important issues, these questions touch upon. Among them are the symbol grounding problem, embodiment and computers/androids making up their own language to cut people out of the communication, just to name a few. Please find it under ‘materials’. “I now boldly state that the symbol grounding problem is solved, by that I mean we now understand enough to create experiments in which groups of agents self-organize symbolic system that are grounded in their interactions with the world and others.” (Steels, 2008)
We are looking forward to an interesting discussion with you!
Your PhenCoCo Team