Google Research & TechTalks

Rob Pike·Google Research & TechTalks

Clarifies concurrency as the composition of independently executing processes (Hoare's CSP) versus parallelism as simultaneous execution, illustrated with live worker-pool topologies.

François Chollet·Google Research & TechTalks

Distinguishes memorized skill interpolation from true sample-efficient skill acquisition on novel tasks via the Abstraction and Reasoning Corpus (ARC).

Rongzhe Wei·Google Research & TechTalks

Proposes a shift from atomic views of knowledge in LLMs toward structured internal knowledge modeling, fundamentally reshaping approaches to unlearning and jailbreaking. Addresses critical trust and safety concerns.

Andy Dong·Google Research & TechTalks

A rigorous examination of privacy amplification techniques for correlated-noise mechanisms, introducing b-min-sep subsampling as a novel batching scheme for DP-MF. Explores advanced topics in private model training.

Connor Wagaman·Google Research & TechTalks

An investigation into local node differential privacy for graphs, presenting a new algorithmic framework for answering linear queries about degree distributions. Offers a deep dive into privacy-preserving graph analysis.

Google Research & TechTalks

This inaugural lecture introduces fundamental concepts of cluster computing and the MapReduce programming model, detailing its architecture and fault tolerance mechanisms for large-scale data processing.