Watching this, on Yannic Kilcher’s transformer channel :


  • Finish E-QiP paperwork – Nope. Gotta fix some things
  • 2:00 Meeting with Vadim.
    • I think we don’t need to determine the sign of the rotation, since there is a single axis of rotation. Going to see if that helps. Nope
    • Just discovered that the deg/rad for get_angle() were reversed! AND THAT FIXED IT
  • Did a little extra bug hunting and added a commanded rate limiter, which needs to be tweaked.

GPT-2 Agents

  • Sim has the Melville model, hopefully there will be some results as well. There are and it looks pretty good. I also had her do some longer texts to see what effect that has. It doesn’t appear to be that much?
  • 3:30 Meeting
    • Work through options for US model. Show the results from the Arabic full tweet probe. One of the neat things about that approach is that the process can be automated as long as there is access to the DB. It will require a way of breaking the results on the training boundary
    • This makes me want to revisit the MD model and see how taking sentences at random could show alternate(?) plot trajectories. Hmm. Not as clear as I thought it would be.
    • Need to start on the Chinese translator/parser
    • Did some interesting word clouds based on the Arabic tweets


  • Move all the chimp stuff into it’s own section. Start to bring in other animal behavior?
  • Work on Moby-Dick. Found a nice summary here: Continued writing. Now at mobydick/section5 in the summary.

DHS PLANXS – Need to review the AI/ML section