Meeting with Xaq, Michael, Chris & Alex
In advance / prep:
- There is this video about Spaun 2.0 from 3 years ago. Ask Chris if this is worth adding as an extra material, would it be helpful for the students or have there been newer developments that will be introduced that could confuse the students?
Meeting notes:
- Xaq gives introduction and explains situation
- Put Neurosymbolic methods into bigger context
- Add it into new content
- Make current material bonus, but replace with cognitive architectures
- If some data changes, whole model changes, can't do interactive (maybe)
- Often take one of the tasks, Q&A, you can get core ability, but doesn't have vision / motor system etc.
- A video showing the run,
- Put information in, ask it questions, would it work?
- Guide interested students onwards to LLMs and the latest results
- Different specialised subsets implement systems, help for generalisation BY BEING SPECIALISED
- Tutorial material, red triangle, blue square,
- Chris: generate random vectors?
- Interactive suggestion:
- two views, one is functional boxes component, other is brain where items are covered
- this is how spaun handles tasks, in the code, input populations, working memory population, binding and unbinding populations, control those to set the memory, work with unbinding populations to ask for explanations
- Binding and unbinding - stays in, Tutorial 1
- Wayson card task to bonus material, actually keep
- Analogies, a lot of that is basic Q&A, we could replace
- DeMorgan's law wrong,
- Section 3 - fix DeMorgan, and operator
Follow-up:
- Send the follow-up, guide students to learn more - Michael mentioned it - ask for link