Releases

Google combines the DeepMind and Google Brain teams into a single unit, Google DeepMind, led by Demis Hassabis.

Meta released DINOv2, a self-supervised vision foundation model and pretrained visual features.

Amazon introduced its first text embedding foundation model through the Bedrock preview.

AWS made its generative-AI coding assistant generally available, including a free individual tier and reference tracking.

AWS announced a managed service for building generative-AI applications with foundation models from Amazon and third parties.

AWS introduced Amazon Titan, its family of text generation and embedding foundation models available through Bedrock.

Databricks released Dolly 2.0 with commercial-use weights and the human-generated databricks-dolly-15k instruction dataset.

CoreWeave publicly detailed its managed inference service for securely serving AI models at scale.

Meta released SAM, its promptable image-segmentation model, together with model weights and the SA-1B dataset.

Baidu launched an ERNIE-powered enterprise coding assistant.

Anthropic launched an early Claude integration for Slack workspaces.

Perplexity launched its native conversational answer-engine application for iPhone and iPad.

Perplexity raised $25.6 million in a Series A led by New Enterprise Associates.

Microsoft introduces Security Copilot, a generative-AI assistant for security investigation, threat intelligence and response.

Baidu AI Cloud launched Qianfan as an enterprise foundation-model and application-development platform.

Databricks released Dolly, a six-billion-parameter instruction-following model that could be trained inexpensively on one machine.

OpenAI introduced plugins connecting ChatGPT to third-party applications and fresh information.

GitHub introduces Copilot X, including conversational Copilot Chat experiences across editors, pull requests and command-line workflows.

Oracle expanded bare-metal GPU instances and high-performance networking for large AI training workloads.

CoreWeave made large-scale NVIDIA H100 Tensor Core GPU infrastructure available for generative-AI training and inference workloads.

Microsoft launches Bing Image Creator, integrating generative image creation into Bing chat and the Edge sidebar.

Google begins opening Bard access in the United States and United Kingdom as an early generative-AI experiment.

NVIDIA launched DGX Cloud, providing browser-based access to managed AI supercomputing infrastructure.

NVIDIA announced broad H100 availability through leading cloud and server providers as generative AI demand surged.

NVIDIA introduced cloud services for enterprises to build custom language and visual generative AI models.

Baidu unveiled its generative conversational assistant, Wenxin Yiyan/ERNIE Bot.

Microsoft introduces Microsoft 365 Copilot and Business Chat, combining large language models with Microsoft Graph and productivity applications.

Google Cloud announces generative AI capabilities in Vertex AI and Generative AI App Builder for enterprise customers.

Anthropic made Claude available through an API and chat interface in its developer console.

Anthropic introduced Claude and Claude Instant more broadly after a closed alpha with partners.

OpenAI released GPT-4, a multimodal large model accepting image and text inputs and producing text outputs.

MosaicML released MosaicBERT, an optimized BERT architecture designed to pretrain high-quality models at much lower cost.

Anthropic described its strategy for building, measuring, and deploying capable AI systems while preparing for potentially transformative impacts.

Microsoft introduces Dynamics 365 Copilot across sales, service, marketing, supply chain and business workflows.

Moonshot AI was founded in Beijing by Yang Zhilin, Zhou Xinyu and Wu Yuxin to develop frontier foundation models.

Cohere established Command as its principal family of enterprise generative language models.

Cohere released Command Light as a lower-latency generation option for enterprise text workloads.

Cohere introduced the Command generation family for enterprise summarization, conversation and text workflows.

OpenAI made the Whisper speech-to-text model available through its API.

OpenAI released the Chat Completions API with gpt-3.5-turbo for developers.

Microsoft Research introduces Kosmos-1, a multimodal large language model capable of perceiving visual inputs and following language instructions.

Meta released the first LLaMA model family for approved research use, spanning 7B to 65B parameters.

AWS and Hugging Face announced a strategic collaboration to make generative AI more accessible and optimize Transformers training and deployment on AWS infrastructure.

Microsoft releases Semantic Kernel to help developers integrate language models, plugins, memory and planning into applications.

Microsoft launches a new AI-powered Bing and Edge with conversational search, synthesized answers and content generation.

Google announces Bard, a LaMDA-powered conversational AI service, for trusted testers ahead of public access.

Google Cloud and Anthropic announce a strategic partnership covering cloud infrastructure, AI accelerators and responsible development of large AI systems.

Anthropic selected Google Cloud as a cloud partner for training and deploying its AI systems, alongside a strategic investment relationship.

OpenAI launched the $20-per-month ChatGPT Plus subscription in the United States.

Google Research introduces MusicLM, a hierarchical text-to-music generation model for high-fidelity audio.

CoreWeave acquired Conductor Technologies and its cloud task-management and burst-rendering platform, extending the cloud into end-to-end VFX and animation workflows.

Microsoft announces a multiyear, multibillion-dollar extension of its OpenAI partnership covering supercomputing, model deployment and Azure commercialization.

OpenAI and Microsoft announced a multiyear, multibillion-dollar extension of their partnership, including Azure supercomputing investment and exclusive cloud-provider status.

Microsoft makes Azure OpenAI Service generally available with enterprise access to OpenAI models through Azure.

Scale announced a restructuring that reduced its employee workforce by roughly 20 percent after expanding too quickly during the prior growth cycle.

Cohere added governed connectors to ground generative applications in enterprise repositories and data systems.

Cohere introduced a conversational Chat API that evolved to support RAG, citations, tools and structured outputs.

Scale launched Outlier to connect subject-matter experts with generative-AI training and evaluation projects.

Hugging Face expanded its organization offering with team and enterprise plans for private repositories, access control, governance, and support.

NVIDIA market capitalisation at end of calendar year 2022, down 51% from the prior year amid a broader semiconductor and tech sector downturn.

Google Research presents Med-PaLM, adapting PaLM to medical question answering and clinical knowledge benchmarks.

Anthropic published Constitutional AI, a method for training a harmless assistant through AI feedback guided by explicit principles.

OpenAI released text-embedding-ada-002, replacing multiple earlier embedding models with a more capable and lower-cost model.

Hugging Face released Parameter-Efficient Fine-Tuning tools for adapting large models with methods such as LoRA.

Perplexity launched its citation-first conversational answer engine for real-time web search.

Magnetar backed CoreWeave with a $100 million strategic investment to expand GPU capacity for AI, machine learning, and visual-computing customers.

OpenAI launched ChatGPT as a conversational research preview.

Meta withdrew the public Galactica demo after criticism over unreliable and misleading generated scientific content.

Meta AI released a research demo of Galactica, a large language model trained on scientific knowledge.

Amazon introduced an AI and computer-vision system able to identify and handle individual products.

BigScience released BLOOMZ and mT0, instruction-finetuned multilingual models derived from BLOOM and the xP3 task collection.

Cohere released its second-generation multilingual embedding model for semantic search across more than 100 languages.

The BigCode project released The Stack, a large permissively licensed source-code dataset with tools for governance and opt-out.

Meta AI published hundreds of millions of predicted metagenomic protein structures with ESMFold.

Cohere established its Rerank family of cross-encoder relevance models for search pipelines.

Cohere made relevance reranking available through a dedicated API endpoint requiring only a small search-stack integration.

Cohere introduced a multilingual cross-encoder Rerank endpoint for improving keyword and semantic search results.

Meta released EnCodec, a neural audio codec for high-fidelity compression.

IBM Research disclosed an AI accelerator built for efficient enterprise inference.

MLflow 2.0 expanded evaluation, pipelines and deployment workflows for production machine learning.

Microsoft announces Designer, an AI-powered graphic-design application using generative image technology.

AWS launched Trainium-powered EC2 instances for high-performance deep-learning training.

Google Research introduces AudioLM for generating long-term coherent, high-fidelity audio from discrete representations.

Hugging Face launched managed, autoscaling dedicated endpoints for deploying models from the Hub.

Meta open-sourced AITemplate, a Python framework that transforms neural networks into optimized GPU code.

Hugging Face released Text Generation Inference, a production server optimized for large-language-model generation.

Meta presented Make-A-Video, a system for generating short videos from text prompts.

OpenAI removed the DALL-E beta waitlist and expanded access to new users.

OpenAI released Whisper, a multilingual speech-recognition model, together with model weights and inference code.

NVIDIA introduced BioNeMo, a generative AI service for chemistry and biomolecular research.

Meta moved PyTorch into an independent foundation hosted by the Linux Foundation.

Perplexity raised approximately $3.1 million in seed financing from Elad Gil and a group of prominent AI and technology investors.

Diffusers added first-class support for Stable Diffusion, enabling local and hosted generation workflows around the newly released open model.

Hugging Face launched its enterprise platform around Private Hub, secure collaboration, governance, deployment, and organizational support.

Hugging Face launched Private Hub as a secure and compliant platform for organizations to collaborate on models, datasets, and applications.

Amazon researchers published a 20-billion-parameter multilingual sequence-to-sequence teacher model.

Baidu launched an ERNIE-powered AI art and creative-generation service.

Hugging Face released Transformer Reinforcement Learning tooling for post-training language models with supervised, preference, and reinforcement-learning methods.

Aravind Srinivas, Denis Yarats, Johnny Ho and Andy Konwinski founded Perplexity in San Francisco.

The AlphaFold database expands to predicted structures for nearly every catalogued protein known to science.