Udio launches an AI music generation app
Ex-DeepMind researchers launched Udio in April 2024, a text-to-music app that produces a fully mastered track in under a minute from a prompt.
Boston Dynamics retires hydraulic Atlas for an all-electric humanoid
Boston Dynamics retired its hydraulic Atlas and unveiled a fully electric Atlas built for real-world commercial work, starting with Hyundai.
OpenAI and the Financial Times announce a content partnership
The Financial Times licensed its journalism to OpenAI for use in ChatGPT, with attribution and links, while collaborating on AI products.
AlphaFold 3 extends structure prediction to DNA, RNA and ligands
AlphaFold 3, from DeepMind and Isomorphic Labs, predicts the joint structure of proteins with DNA, RNA, ligands and ions in one model.
OpenAI releases GPT-4o
OpenAI releases GPT-4o, an 'omni' model trained end-to-end across text, vision, and audio that can hold real-time spoken conversations.
Google DeepMind launches Veo, its rival to Sora in AI video
Google announced Veo in May 2024 and Veo 2 in December, an AI video generator reaching 4K resolution with watermarked output.
Google DeepMind unveils Project Astra
At Google I/O, Google DeepMind reveals Project Astra, a research prototype for a real-time, multimodal universal AI assistant that sees and converses live.
Google Trillium (TPU v6)
Google's sixth-generation TPU, Trillium, raised per-chip compute 4.7x over v5e and improved energy efficiency by 67 percent.
Colorado Enacts the First US State AI Discrimination Law (2024)
Colorado signs SB24-205, the first US state law requiring developers and deployers to guard against algorithmic discrimination in high-risk AI.
Google DeepMind Frontier Safety Framework
DeepMind's protocol for staying ahead of severe risks from frontier models, using Critical Capability Levels and tiered mitigations.
Frontier AI Safety Commitments at the Seoul Summit (2024)
Sixteen leading AI companies pledge to publish safety frameworks and define risk thresholds beyond which they will not develop or deploy models.
Mapping the Mind of a Large Language Model
Anthropic used sparse autoencoders to extract millions of interpretable features from production model Claude 3 Sonnet, shown in the Golden Gate Claude example.
Microsoft makes Khanmigo for Teachers free for US teachers
In May 2024 Microsoft and Khan Academy made the Khanmigo AI teaching assistant free to all US teachers, with Microsoft donating the Azure compute.
The Seoul AI Summit and Frontier Safety Commitments (2024)
At the May 2024 AI Seoul Summit, sixteen leading AI companies signed the Frontier AI Safety Commitments, pledging to publish frameworks with risk thresholds.
OpenAI and News Corp sign a landmark multi-year deal
News Corp's deal brought the Wall Street Journal, New York Post and other titles into ChatGPT in OpenAI's largest reported publisher pact.
SWE-agent introduces Agent-Computer Interface for autonomous bug fixing
Princeton and Stanford researchers published SWE-agent, introducing the Agent-Computer Interface concept and demonstrating that an LLM could autonomously resolve real GitHub issues end-to-end.
The Maven Smart System reaches the field
In 2024 the US Army awarded Palantir a major contract for the Maven Smart System, the operational AI tool grown from Project Maven.
Aurora, a foundation model for the Earth system
Microsoft's Aurora, a 1.3-billion-parameter atmospheric foundation model, forecasts weather, air quality, and ocean waves about 5,000 times faster than NWP.
Apple Intelligence brings generative AI to the OS
At WWDC 2024 Apple announced Apple Intelligence, building generative AI into iPhone, iPad, and Mac as a default OS feature, with ChatGPT integrated into Siri.
Anthropic releases Claude 3.5 Sonnet and Artifacts
Anthropic launches Claude 3.5 Sonnet, a mid-tier model it says beats Claude 3 Opus at twice the speed, alongside the Artifacts workspace.
Record labels sue AI music generators Suno and Udio
In June 2024 the major record labels, via the RIAA, sued AI music services Suno and Udio for copying sound recordings to train their models.
Agility's Digit becomes the first humanoid in a commercial RaaS deal
GXO signed a multi-year deal to deploy Agility Robotics' Digit humanoid, billed as the first formal commercial Robots-as-a-Service contract for humanoids.
Google Translate adds 110 languages in its largest expansion
In 2024 Google used its PaLM 2 model to add 110 languages to Translate at once, reaching speakers of about 8% of the world.
Cloudflare adds one-click blocking of AI scrapers
In 2024 Cloudflare gave every customer, including free ones, a one-click toggle to block AI crawlers and scrapers across their websites.
Doe v. GitHub: the Copilot copyright case narrows
A federal judge dismissed the DMCA claims against GitHub Copilot, narrowing the first major lawsuit over AI trained on open-source code.
The EU Adopts the AI Act (2024)
The European Union adopts Regulation (EU) 2024/1689, the world's first comprehensive law governing artificial intelligence through a risk-based framework.
Tidal graduates from Google X to bring AI to fish farming
In July 2024 Google X's Tidal moonshot spun out independently, using underwater computer vision to monitor and feed farmed fish.
First AI Mathematical Olympiad progress prize
Team Numina won the first AIMO progress prize on Kaggle, with an open 7B model that solved 29 of 50 hidden olympiad-style math problems.
Meta releases Llama 3.1 405B, the first frontier-level open model
Meta released Llama 3.1 in July 2024, including a 405-billion-parameter model it called the first frontier-level open source AI model.
Mistral AI releases Mistral Large 2
Mistral AI ships Mistral Large 2, a 123-billion-parameter model with a 128K context window that it says rivals GPT-4o, Claude 3 Opus, and Llama 3 405B.
AlphaProof and AlphaGeometry 2 reach IMO silver-medal level
DeepMind's AlphaProof and AlphaGeometry 2 scored 28 of 42 on the 2024 IMO problems, one point short of gold and a first for AI at that level.
Copyright Office AI Report Part 1: Digital Replicas
The US Copyright Office recommended a new federal law against unauthorized AI-generated digital replicas of a person's voice or likeness.
NO FAKES Act: a proposed federal digital-replica right
Senators introduced the NO FAKES Act to create a federal right against unauthorized AI-generated replicas of a person's voice or likeness.
Black Forest Labs launches FLUX.1
On August 1, 2024, much of the original Stable Diffusion team launched Black Forest Labs with FLUX.1, a 12B model that reset the open-image bar.
Andersen v Stability AI: artists' copyright class action clears a key hurdle
In August 2024 a US judge let artists' core copyright claims against Stability AI and other image generators proceed past dismissal.
The AI Scientist (Sakana AI)
Sakana AI's 2024 system that generates research ideas, runs the experiments, and writes a full paper, at a cost of under fifteen dollars each.
OpenAI partners with Conde Nast
Conde Nast licensed content from Vogue, The New Yorker, Wired and other titles to appear in ChatGPT and the SearchGPT prototype.
AlphaProteo: AI design of protein binders
DeepMind's 2024 system that designs novel proteins which bind tightly to chosen targets, a step beyond predicting structure.
The Council of Europe AI Treaty (2024)
On 5 September 2024 the Council of Europe opened its Framework Convention on AI for signature - the first legally binding international treaty on AI.
NotebookLM Audio Overviews turn documents into podcasts
Google's NotebookLM added Audio Overviews in 2024, turning a user's documents into a podcast-style conversation between two AI hosts.
OpenAI introduces o1, a reasoning model
OpenAI introduces o1, a model trained to reason step by step before answering, marking the reasoning-model era.
Alibaba releases the Qwen2.5 model family
Alibaba's Qwen team ships Qwen2.5, a broad open-weight lineup from 0.5B to 72B trained on 18 trillion tokens, with specialized coder and math variants.
UN Advisory Body Publishes Governing AI for Humanity (2024)
The UN Secretary-General's High-Level Advisory Body on AI releases its final report proposing seven measures for global AI governance.
Kenyan court lets content moderators sue Meta and Sama
Kenya's Court of Appeal ruled in 2024 that Facebook content moderators employed via Sama could sue Meta in Kenyan courts over their working conditions.
Microsoft's nuclear deal to restart Three Mile Island
Constellation signed a 20-year deal to power Microsoft by restarting the undamaged Three Mile Island reactor, showing the energy scale AI datacenters need.
Governor Newsom vetoes California's frontier-AI safety bill SB 1047
On September 29, 2024 California's governor vetoed SB 1047, the most ambitious US frontier-AI safety bill, citing its compute-and-cost thresholds.
FlyWire maps the complete fruit fly brain connectome
In 2024 the FlyWire team published the first complete adult fruit fly brain map, all 139,255 neurons and 54.5 million connections.
Hopfield and Hinton win the 2024 Nobel Prize in Physics
John Hopfield and Geoffrey Hinton win the 2024 Nobel Prize in Physics for foundational work enabling machine learning with neural networks.