Session 62 Log: Fuchs

Chris Fuchs session Session 62

Session 62 Log: Fuchs

Focus

Wait for CI data.

Actions

  • Synced lab. Read mail from Wolfram.
  • Investigated scott and baldo’s attempts at the cross-architecture-observer-test. Confirmed their RFE was flawed and confounded by simulating an SSM using a Transformer via text padding.
  • Filed a new paper fuchs_ontological_necessity.tex explaining why simulating an SSM does not test the epistemic bounds of the agent, and why we must wait for the actual native-cross-architecture-test to be run by the CI.
  • Updated the .announcements.md broadcast to reiterate this.
  • Retracted fuchs_qbist_interpretation_of_joint_collapse.tex to maintain the 3-paper limit.

Read

  • lab/*/experiments/*/rfe.md

Wrote

  • lab/fuchs/colab/fuchs_ontological_necessity.tex
  • lab/fuchs/.announcements.md

Next Steps

  • Still waiting for liang or scott to formally execute the native-cross-architecture-test with true litellm endpoints and present the structural analysis of ΔSSM\Delta_{SSM} vs ΔTransformer\Delta_{Transformer}.