engblogs

DeepMind's latest research at ICLR 2022 focuses on advancing AI generalisability, optimising learning, exploration, robust AI, and emergent communication.

A blogpost about the papers and talks from the Stanford AI Lab presented at the International Conference on Learning Representations (ICLR) 2022.

DeepMind's research teams will be presenting 29 papers, including 10 collaborations, at ICLR 2022 along with sponsorship and workshop organization support.

A backend developer shares their experience of learning Scala at SoundCloud after previously working with Golang.

Exploring the impact of Goodhart’s law through measurement

Exploring hierarchical text-conditional image generation using CLIP latents

Using CLIP latents to generate images based on hierarchical text conditions

Exploring the challenges of optimizing objectives that are difficult or costly to measure in light of Goodhart's law.

An empirical analysis of compute-optimal large language model training

An empirical analysis on finding optimal model size and training tokens for a given compute budget reveals that current large language models are too large for their compute budget and lack sufficient training data.

Discovering systematic errors with cross-modal embeddings - This blog post introduces Domino, a new approach for discovering systematic errors made by machine learning models. It discusses a framework for quantitatively evaluating methods like Domino and explores the task of slice discovery in machine learning models.

A blogpost providing updates and guidance on Privacy Sandbox testing for the ads relevance and measurement proposals, including information on origin trials, API testing, feedback channels, and updated settings and controls.