
Learning the language of molecules to predict their properties
An AI system predicts molecular properties with limited data, accelerating drug discovery and material development.

Modular visual question answering via code generation
The blogpost discusses a modular visual question answering framework called CodeVQA that uses code generation to answer visual questions using program synthesis, improving accuracy on complex reasoning tasks and outperforming prior few-shot methods.

Association for Computational Linguistics (ACL) 2023
Apple-sponsored workshops and events at the Association for Computational Linguistics (ACL) 2023 conference in Toronto, Canada.
Why I chose Duolingo: Jackie Lee, AI Research Scientist
A firsthand account from Jackie Lee, an AI Research Scientist, on why she chose to work at Duolingo.

How NVIDIA Omniverse bolsters AI with synthetic data
Discover how NVIDIA Omniverse enhances AI by leveraging synthetic data for training perception models.

Frontier AI regulation: Managing emerging risks to public safety
Navigating regulatory challenges in the realm of AI to safeguard public safety

Pic2Word: Mapping pictures to words for zero-shot composed image retrieval
Mapping pictures to words for zero-shot composed image retrieval using Pic2Word method

How NVIDIA Omniverse bolsters AI with synthetic data
The blog post explores how NVIDIA's Omniverse enhances AI by utilizing synthetic data for training perception models.

Frontier AI regulation: Managing emerging risks to public safety
Managing emerging risks to public safety in the context of Frontier AI regulation.

We're sponsoring SoME3
Jane Street announces sponsorship of SoME3, a contest promoting development of math education videos.

MIT scientists build a system that can generate AI models for biology research
A description of BioAutoMATED, an open-source, automated machine-learning platform designed to generate AI models for biology research and foster the democratization of artificial intelligence in research labs.

Pic2Word: Mapping pictures to words for zero-shot composed image retrieval
This blogpost discusses Pic2Word, a method for mapping pictures to words for zero-shot composed image retrieval, enabling precise retrieval of target images by combining image and text.