Causal Inference in Deep Learning (Podcast Overview with Brady Neal)

[@LiaAnggraini1]

thank you! been waiting for this kinda topic😄

[@connor-shorten]

Really glad you enjoyed it, thanks for watching!

[@Ceelvain]

At last! I really think causal inference is one of the crucial building blocks to bridge the gap between symbolic AI and neural networks. And therefore a crucial step toward AGI.

[@connor-shorten]

Definitely! In addition ton the future promise, it seems like there are a lot of "low hanging fruit" research ideas to pursue as well in the integration of Causal Inference and Deep Learning!

[@anishbhanushali]

I'm a bit sceptical about causal inference + deep learning. Since it looks like re-packaging of the older nomenclature. Causal inference looks very similar to what PGM(probabilistic graphical model) used to represent, I could be wrong though ... may an in-depth analysis could reveal its novelty.

[@connor-shorten]

What do you think about using Deep Learning for similarity analysis / interchangeability in Causal Inference?

[@anishbhanushali]

@@connor-shorten I see 2 ways in which DL can mix very well with causal inference . 1) Using it's generative ability (GANs, AutoEnc etc .. ) to extrapolate examples for multiple causes. 2) Providing higher level input variables to classical causal inference pipelines for an example "what adv video clip can cause more interest in product inquiry?" ... to extract high level variables like "number of people", "pet or no pet", "type of background", "music to speech ratio" etc.. we need DL solution. Cherry on top if E2E pipeline is differentiable

[@connor-shorten]

Please check out the full podcast with Brady Neal! kzbin.info/www/bejne/qmjKaqZnfLyWeJo