4 ms·
I think that argument ignores that neural nets can learn the models that Pearl believes are important. And it's quite plausible that human reasoning works in s
by fjsolwmv 9y ago
I think that argument ignores that neural nets can learn the models that Pearl believes are important. And it's quite plausible that human reasoning works in similar way.
Pearl's career is all about causal networks, so he's slightly biased when it comes to survey like this.
- vokolk 9y agoSo you are saying that Pearl's argument that current ML system cannot reason about interventions and retrospection (as he defines them) are wrong?