Neural Network
11 articles about Neural Network
Understanding neural networks using sparse circuit models
OpenAI is using sparse circuit models to better interpret how neural networks reason, aiming for more transparent, safer, and more reliable AI.
Introducing Triton 1.0: open-source GPU programming language for neural networks
Triton 1.0 is an open-source, Python-like language that lets researchers write near-expert-level GPU code for neural networks without CUDA.
Analysis shows compute efficiency for training neural networks improves faster than Moore’s Law since 2012
An analysis finds that since 2012, algorithmic advances have cut the compute needed for ImageNet-level neural nets by 2x every 16 months—about 44x less than AlexNet—outpacing hardware gains.
OpenAI Five neural networks begin defeating amateur teams in Dota 2
OpenAI Five, a team of five neural networks, is starting to beat amateur human teams at Dota 2.
Open-source neural network Whisper achieves near human-level English speech recognition
OpenAI introduced and open-sourced Whisper, a neural network for English speech recognition with near human-level accuracy and robustness.
CLIP neural network connects text and images for zero-shot visual classification
CLIP is a neural network that learns visual concepts from text and can classify images in a zero-shot way using only category names.
Jukebox: a neural network that generates music and singing in various styles
Jukebox is a neural network that generates raw-audio music (including basic singing) in many genres and artist styles, with released code, weights, and a sample-exploration tool.
Using neural networks to assign types for entity disambiguation
A neural network disambiguates what a word refers to by checking which of ~100 automatically discovered, non-exclusive types it fits.
Training a neural network to play Minecraft using video pretraining
A neural network was trained via video pretraining on human Minecraft gameplay to use keyboard/mouse and, with minimal labeled data, learn complex tasks like crafting diamond tools—advancing toward general computer-using agents.
DALL·E: neural network generating images from text captions
DALL·E is a neural network that generates images from natural-language text captions.
Neural networks enable a robot hand to solve the Rubik’s Cube
Neural networks trained in simulation use reinforcement learning and domain randomization to control a robot hand that solves a Rubik’s Cube and adapts to unexpected real-world disturbances.
Showing page 1 of 1