Few issues today generate as much fascination, worry, and confusion as Artificial Intelligence (AI).

The first week of the 35th conference on Neural Information Processing Systems (NeurIPS2021) saw eight fascinating invited talks, tutorials, affinity group workshops, and a new datasets and benchmarks track. There were also poster sessions, oral sessions, competitions, demonstrations, and more. With this compilation of tweets, we look back on the week. "The greatest violence is the product of remoteness from reality" – a great talk by Mary L. Gray, The Banality of Scale: A Theory on the Limits of Modeling Bias and Fairness Frameworks for Social Justice (and other lessons from the Pandemic) at #NeurIPS2021 'How duolingo uses AI to Asses, Engage and Teach Better' session @NeurIPSConf is . The invited talk by @lugosi_gabor at #NeurIPS2021 was very enjoyable and also a little disturbing -- if your data isn't Gaussian (or subgaussian), then even basic things become nontrivial.https://t.co/TPByspYh5T The final #NeurIPS2021 keynote starts soon!

A principal one among them is the existence of multiple domains that share the same underlying causal structure for actions. We describe an approach that exploits this shared causal structure to discover a hierarchical task structure in a source domain, which in turn speeds up learning of task execution knowledge in a new target domain. Our approach is theoretically justified and compares favorably to manually designed task hierarchies in learning efficiency in the target domain. We demonstrate that causally motivated task hierarchies transfer more robustly than other kinds of detailed knowledge that depend on the idiosyncrasies of the source domain and are hence less transferable. These domains are complex, and good performance requires selecting long chains of actions to achieve subgoals needed for ultimate success.

Sequential decision tasks present many opportunities for the study of transfer learning. A principal one among them is the existence of multiple domains that share the same underlying causal structure for actions. We describe an approach that exploits this shared causal structure to discover a hierarchical task structure in a source domain, which in turn speeds up learning of task execution knowledge in a new target domain. Our approach is theoretically justified and compares favorably to manually designed task hierarchies in learning efficiency in the target domain. We demonstrate that causally motivated task hierarchies transfer more robustly than other kinds of detailed knowledge that depend on the idiosyncrasies of the source domain and are hence less transferable.