OpenAI opens the DALL·E 2 API in public beta
OpenAI makes DALL·E image generation and editing available to all API customers for integration into applications and products.
EVENT ARCHIVE
Events
English reports are listed here. AI-generated drafts are clearly labeled until an editor reviews them.
AI-assisted · reviewedImpact 93/100
Google releases open Flan-T5 checkpoints
Google instruction-tuned the T5 family on 1,836 tasks and released five checkpoints ranging from 80 million to 11 billion parameters.
DALL·E 2 drops its waitlist and opens immediate sign-up
OpenAI removes the DALL·E beta waitlist, allowing users to sign up immediately; more than 1.5M active users were generating over 2M images per day.
OpenAI releases Whisper
OpenAI released and open-sourced Whisper, an automatic speech recognition model for multilingual transcription, language identification, and speech translation into English.
Stable Diffusion becomes publicly available
Stability AI publicly releases Stable Diffusion, bringing an open-weight image-generation model to developers and creator communities.
Midjourney enters public testing
Midjourney opens public testing and provides a low-barrier text-to-image service through Discord.
DeepMind introduces the Flamingo visual language model
DeepMind introduced Flamingo, an 80-billion-parameter visual language model that used a few examples to handle image, video, and text tasks.
OpenAI introduces DALL·E 2 in a limited research preview
OpenAI introduces DALL·E 2, built around CLIP image embeddings and a diffusion decoder, and initially opens it only to a small group of invited users.
DeepMind introduces the competitive programming system AlphaCode
DeepMind introduced the code-generation system AlphaCode, which performed at about the median level of participants in Codeforces competitions.
DeepMind introduces the 280-billion-parameter Gopher language model
DeepMind introduced Gopher, a 280-billion-parameter language model, and published research on its capabilities, limitations, and risks without releasing public weights or an API.
OpenAI releases Codex
OpenAI releases a code-generation model fine-tuned from GPT-3 and used in the early GitHub Copilot.
EleutherAI releases GPT-Neo
EleutherAI releases an early open alternative to GPT-3.
Google publishes Switch Transformer, scaling a sparse model to 1.6T parameters
Switch Transformer routed each token to a single expert, scaling total sparse parameters to about 1.6 trillion without activating the full model for every token.
OpenAI introduces the 12B text-to-image model DALL·E
OpenAI introduces DALL·E, a 12B autoregressive Transformer that models text and discrete image tokens as one sequence.
OpenAI releases the CLIP vision-language model
OpenAI released CLIP, a vision-language model trained with natural-language supervision, together with its paper, code, and pretrained models.
Meta publishes BART research and pretrained models
Meta published the BART denoising sequence-to-sequence model and made code and pretrained weights available through fairseq.
GShard paper demonstrates a sparse-expert translation model with more than 600B parameters
Google researchers introduce GShard automatic sharding and use it to train a multilingual translation MoE with more than 600B total parameters on 2,048 TPU v3 accelerators.
OpenAI opens GPT-3 access through a private API beta
OpenAI launches a private beta for a general-purpose text-in, text-out API running GPT-3 family models, offering controlled access instead of open weights.
GPT-3 demonstrates in-context learning at 175B parameters
OpenAI publishes GPT-3 research showing that a 175B-parameter model can switch tasks from instructions and a few examples in the prompt without updating its weights.
Google publishes T5 and releases models, code, and C4
T5 cast a wide range of natural-language-processing tasks into a text-to-text format and released a model family ranging from 60 million to 11 billion parameters.