Reflections:
- Hossenfelder talks about recent research looking at how scientist think large language models think or use reasoning and she makes the case that there is no evidence that the model is thinking or aware of itself
- Even maths calculations end up being the model using prediction tokens in a really roundabout way to get to an answer that is ‘correct’
- The research is very much against the idea that LLMs are conscious and go against the hype that is pushed in public popular media
- My takeaway is that I need to remain critical when trying to make sense of the claims of AI companies about how their models are working what they can actually do
- At the moment – it’s probably marketing b.s…
Reference
Sabine Hossenfelder (2025) New Research Reveals How AI ‘Thinks’ (It Doesn’t), YouTube. Available at: https://www.youtube.com/watch?v=-wzOetb-D3w (Accessed: 13 May 2025).
