
DALL·E 2 pre-training mitigations
Exploring strategies to improve pre-training efficiency for advanced AI model DALL·E 2

DALL·E 2 pre-training mitigations
Mitigations implemented to reduce risks associated with powerful image generation models.

Learning to play Minecraft with Video PreTraining
Enhancing Minecraft gameplay through Video PreTraining techniques.

Leading a movement to strengthen machine learning in Africa
Leading a movement to strengthen machine learning in Africa: Avishkar Bhoopchand, a research engineer, shares his journey to DeepMind and his involvement in the Deep Learning Indaba to raise the profile of deep learning across Africa.

Chrome 104 Beta: New Media Query Syntax, Region Capture, and More
The blogpost introduces the new features and improvements in Chrome 104 Beta, including region capture, updated media query syntax, and new origin trials.

Learning to play Minecraft with Video PreTraining
Training a neural network with Video PreTraining to play Minecraft using unlabeled video dataset and small amount of labeled contractor data, achieving proficiency in crafting diamond tools that usually takes over 20 minutes for humans.

Leading a movement to strengthen machine learning in Africa
Promoting and advancing the field of machine learning in Africa through a comprehensive movement

Microsoft’s framework for building AI systems responsibly
Building AI systems responsibly with Microsoft's framework

BYOL-Explore: Exploration with Bootstrapped Prediction
BYOL-Explore is a curiosity-driven AI agent that learns a world representation, world dynamics, and exploration policy together by optimizing prediction error at the representation level.

BYOL-Explore: Exploration with Bootstrapped Prediction
A blogpost on BYOL-Explore, a method for curiosity-driven exploration in visually-complex environments by optimizing a single prediction loss in the latent space without auxiliary objectives.

Evolution through large models
Exploring advanced machine learning techniques with large models

Unlocking High-Accuracy Differentially Private Image Classification through Scale
Unlocking high-accuracy differentially private image classification through scale.