integrating POVs, using sheaves

[timelets]

Watch on YouTube

Comments

[dmm]

Спасибо!

Очень интересно (утверждение, что "disagreement is handled by cohomology", this might be the missing piece for me).

[dmm]

Oh, and the author is super-interesting, he has this draft book: https://www.davidjaz.com/Papers/DynamicalBook.pdf

[timelets]

Yes, he's quite good. He's into applying CT and cohomology to systems thinking, which I really like.

Thanks for the book link!

[dmm]

Книжка упоминает когомологии только в одном месте, но очень интересным образом:

pp. 253-254:

"If instead of sets we were working with homotopy types, then
level 0 would still be elements, and level −1 would be identifications between these
elements, and −2 identifications between these identifications, and so on. In general,
the negative levels would have (increasingly abelian) cohomological information about
the positive levels."

[timelets]

Thanks. Speaking of cohomology, imagine we have a system of agents, both human and artificial, that talk to various AI services. Inevitably, they run into contradictions as described in the video. A cohomology approach seems to be a way forward in resolving conflicts at the system level.
Now, human and artificial agents run on different time scales. Aligning them just on that is an organizational and technical problem that has to be solved for the system to work.

[dmm]

Thanks. Yes, time scale differences are likely to be particularly non-trivial. (I suspect that those differences which are well captured by the cohomological approach are "spatial" in a generalized sense of the word, and generalization to temporal aspects would need further non-trivial work.)