AI News

In machine learning, synthetic data can offer real performance improvements

Models trained on synthetic data can be more accurate than other models in some cases, which could eliminate some privacy, copyright, and ethical concerns from using real data.

Study urges caution when comparing neural networks to the brain

Computing systems that appear to generate brain-like activity may be the result of researchers guiding them to a specific outcome.

Machine learning facilitates “turbulence tracking” in fusion reactors

A new approach sheds light on the behavior of turbulent structures that can affect the energy generated during fusion reactions, with implications for reactor design.

Using sound to model the world

This machine-learning system can simulate how a listener would hear a sound from any point in a room.

Scaling laws for reward model overoptimization

Introducing Whisper

We’ve trained and are open-sourcing a neural net called Whisper that approaches human level robustness and accuracy on English speech recognition.

Efficient training of language models to fill in the middle

A hazard analysis framework for code synthesis large language models

DALL·E 2 pre-training mitigations

In order to share the magic of DALL·E 2 with a broad audience, we needed to reduce the risks associated with powerful image generation models. To this end, we put various guardrails in place to prevent generated images from violating our content policy.

Learning to play Minecraft with Video PreTraining

We trained a neural network to play Minecraft by Video PreTraining (VPT) on a massive unlabeled video dataset of human Minecraft play, while using only a small amount of labeled contractor data. With fine-tuning, our model can learn to craft diamond tools, a task that usually takes proficient humans over 20 minutes (24,000 actions). Our model uses the native human interface of keypresses and mouse movements, making it quite general, and represents a step towards general computer-using agents.