GenAI Chronicle
2010
2011
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2023

December

5 events

Editors' summary

Google answers with Gemini, the EU agrees on AI rules

On 6 December, Google announced Gemini 1.0, replacing Bard. It came in three sizes: Ultra, Pro and Nano.

On the 8th, the EU reached political agreement on the AI Act, the first comprehensive AI regulation anywhere.

On the 11th, France's Mistral AI released Mixtral 8x7B and raised $415 million alongside it. It drew attention as a serious option among open models.

On the 27th, the New York Times sued OpenAI and Microsoft for copyright infringement.

This block is written by the editors. It is kept separate from the sourced record below.

Record5 events
  1. 6
    ModelGoogle / Sundar Pichai / Demis Hassabis

    Google announces Gemini 1.0 in Ultra, Pro, and Nano

    Sundar Pichai announced Gemini 1.0, the first major result from the merged Google DeepMind and built to be natively multimodal — handling text, code, audio, images, and video as one from the start. It came in three sizes, Ultra, Pro, and Nano, with Nano intended to run on device, and Pro went into Bard the same day. As the third generation of foundation model after LaMDA and PaLM 2, it also became the name Google moved its products toward. The promotional video meant to show its abilities drew criticism: it was not a real-time exchange but an edited result obtained from still frames pulled out of footage plus text prompts. Google said the results were genuine but that "latency has been reduced and Gemini outputs have been shortened for brevity." After February's wrong answer from Bard, Google was again being asked about how it staged a demo.

  2. 8
    PolicyEuropean Union / European Parliament / Council of the European Union

    EU reaches political agreement on the AI Act

    The European Parliament and the Council of the EU reached political agreement on the AI Act after three days of negotiation. The framework scales obligations to risk: it bans practices such as social scoring, emotion recognition in workplaces and schools, and biometric categorisation using sensitive traits. General-purpose AI models must keep technical documentation and comply with EU copyright law, while high-impact models with systemic risk face additional duties — model evaluation, adversarial testing and reporting of serious incidents. Fines run to 7% of global turnover or 35 million euro. It is the first comprehensive AI regulation anywhere. The deal was struck late on 8 December; both institutions announced it on the 9th.

  3. 11
    ModelMistral AI

    Mixtral 8x7B brings a sparse mixture of experts to open weights

    Mistral AI released Mixtral 8x7B, a sparse mixture-of-experts model with eight experts: roughly 46.7 billion parameters in total but only about 12.9 billion used per token. It beat Llama 2 70B on most benchmarks with six times faster inference, and shipped under the same Apache 2.0 licence as Mistral 7B. The significance was that a technique the large labs had been using without describing was now available as weights anyone could download; Google would not confirm MoE in a flagship model until Gemini 1.5 Pro in February 2024. The same day, Mistral disclosed a $415 million (€385 million) Series A led by Andreessen Horowitz at a valuation of about $2 billion, and opened a commercial platform serving both Mistral 7B and Mixtral through an API.

  4. 27
    PolicyOpenAI / Microsoft / The New York Times

    New York Times sues OpenAI and Microsoft for copyright infringement

    The New York Times sued OpenAI and Microsoft in the US District Court for the Southern District of New York, alleging that millions of its articles were used without permission to train AI models and that products like ChatGPT could reproduce portions of articles nearly verbatim, competing with its journalism. It sought billions of dollars in damages. The first copyright suit against AI companies by a major news organization, it became a defining legal battle over training data and fair use in generative AI.

  5. 29
    ProductMicrosoft / OpenAI

    GitHub Copilot Chat becomes generally available in all plans

    GitHub, a Microsoft subsidiary, made GitHub Copilot Chat generally available, letting developers work through an AI conversation inside the editor. It ran in Visual Studio Code and Visual Studio and was powered by GPT-4. Chat was included in every existing GitHub Copilot plan at no extra cost, switched on automatically for individual subscribers and enabled by administrators for organizations. Verified teachers, students, and maintainers of popular open source projects could use it free. The feature had been shown as a technical preview under Copilot X in March, nine months earlier.

↑ 2023