Visit ai.facebook.com → On Hacker News Apr 2019 – Jun 2023
AI & Machine Learning 27Tools & Productivity 2Software Engineering 2Math & Algorithms 2Games & Graphics 2Databases & Data 2
Meta AI has built CICERO, the first AI to achieve human-level performance in the popular strategy game Diplomacy. It's a breakthrough toward building AI that can use language to work with people to achieve strategic goals.
Today, we’re releasing our LLaMA (Large Language Model Meta AI) foundational model with a gated release. LLaMA is more efficient and competitive with previously published models of a similar size on existing benchmarks.
Aroma is a code-to-code search and recommendation tool that uses ML to make the process of gaining insights from big codebases much easier.
We have open-sourced wav2letter@anywhere, an inference framework for online speech recognition that delivers state-of-the-art performance.
Voicebox is a state-of-the-art speech generative model based on a new method proposed by Meta AI called Flow Matching. By learning to solve a text-guided speech infilling task with a large scale of data, Voicebox outperforms single-purpose AI models across speech tasks through in-context learning.
To enable speech recognition technology for many more languages spoken around the globe, Facebook AI is releasing wav2vec Unsupervised, a new method to train models with no supervision whatsoever. Wav2vec Unsupervised rivals the performance of the best supervised systems from just a few years ago.
Our work aims to break down language barriers across the world for everyone to understand and communicate with anyone—no matter what language they speak.
How can we build machines with human-level intelligence? There’s a limit to how far the field of AI can go with supervised learning alone. Here's why self-supervised learning is one of the most promising ways to make significant progress in AI.
TransCoder is the first self-supervised neural transcompiler system for migrating code between programming languages. It can translate code from Python to C++, for example, and it outperforms rule-based translation programs.
Facebook AI has developed a new technique to mark the images in a dataset, so that researchers can then determine if a particular machine learning model...
Text-to-speech systems across the industry typically rely on GPUs or specialized hardware to generate state-of-the-art speech in real-time production. We solved core efficiency challenges to process one second of audio in 500 milliseconds — using only CPUs.
TaBERT is the first model that has been pretrained to learn representations for both natural language sentences and tabular data.
Deep and machine learning (ML) frameworks are good at what they do, but altering their underlying code is typically difficult. Flashlight is a new open source library written entirely in C++ that enables teams to rapidly modify frameworks to fit their needs. Read more:
We created data2vec 2.0, a general self-supervised AI algorithm for speech, vision, and text that can train models 16x faster than the most popular existing algorithm for images while achieving the same accuracy.
Deep neural networks have long been considered too complex to understand from first principles — but new research does just that, presenting a theoretical framework for DNNs.
We are open-sourcing BoTorch and Ax, two new tools that leverage adaptive experimentation to efficiently tackle optimization challenges in tuning software configurations, machine learning hyperparameters, ranking systems, and more.
We are releasing HiPlot, a lightweight interactive visualization tool to help AI researchers discover correlations and patterns in high-dimensional data.
Sending an MP3 typically requires 128 kb/s of bandwidth. We can compress HiFi audio down to 12 kb/s, without sacrificing the quality.
As part of the fastMRI research project to use AI to speed up MRI scans, NYU Langone Health is making a new dataset of de-identified brain MRIs available to researchers and Facebook AI is sharing additional tools and resources.
To advance research on AI that can understand others’ points of view and collaborate effectively, Facebook AI has developed a bot that sets a new state of the art in Hanabi, a card game in which all players work together.
Facebook AI has made scientific progress in improving nuanced conversational skills, including consistency, specificity, and empathy.
We’re announcing that Meta AI has built and released BlenderBot 3, the first 175B parameter, publicly available chatbot complete with model weights, code, datasets, and model cards. We’ve deployed it in a live interactive conversational AI demo.
Does your child love to draw? Ever wished that characters in their drawings could “come to life” and move around the page? Using AI, we’ve developed automatic animation that can bring children’s one-of-a-kind characters to life!
In 2020, we initiated the Meta Training and Inference Accelerator (MTIA) family of chips to support our evolving AI workloads, starting with an inference accelerator ASIC for deep learning recommendation models (DLRMs).
Facebook AI is introducing M2M-100, the first multilingual machine translation model that can translate between any pair of 100 languages without relying on English data.
Facebook AI has developed the first neural network that uses symbolic reasoning to solve advanced mathematics problems.