2023: A brief AI recap of the year
2023 was a huge year for AI, with many groundbreaking developments. And what better way to start 2024 than by reviewing the major AI milestones from last year as we prepare for new ML-powered advancements. So, without further ado, here they are, listed by the month:
January
Microsoft partners with OpenAI, investing $10 billion into the company, with the end goal of integrating OpenAI’s technology into its ecosystem, namely Microsoft Edge and Bing.
February
Meta releases LLaMA, its own large language model (LLM) that boasts 65 billion parameters.
Utilizing ChatGPT technology, Microsoft launches Bing Chat (later renamed Microsoft Copilot) — the first generative AI solution to offer real-time internet access.
March
OpenAI releases GPT-4, making it available to Plus subscribers. The multimodal LLM, boasting 1.76 trillion parameters, accepts both image and text inputs.
On the very same day, Anthropic releases Claude — a new AI chatbot trained on the company’s namesake LLM.
Google begins to offer early access to Bard, its generative AI chatbot, powered by the company’s LLM known as LaMDA. The bot is available for testing to users in the US and the UK.
Meta’s LLaMA suddenly gets leaked as a downloadable torrent.
April
Hugging Chat is released to the general public by Hugging Face in an effort to offer users “the first open source alternative to ChatGPT.” The generative AI bot is based on Meta’s LLaMA leaked online a month earlier.
May
Google announces the release of PaLM2, its latest LLM trained on 340 billion parameters, which soon gets integrated into Bard.
Anthropic boosts Claude, which can now summarize entire books in seconds, reportedly making it the most powerful AI summarization tool on the market.
A viral video featuring US Vice President Kamala Harris is exposed as a deepfake, marking the second major incident of this kind following a fabricated image of Pope Francis wearing a puffer jacket. Integrity concerns about AI-generated imagery are voiced in the news.
Yandex introduces its proprietary LLM named YandexGPT, equipped with 100 billion parameters, subsequently integrating the model into its flagship virtual assistant, Alice. This becomes the very first generative AI solution from Russia.
June
Apple introduces Apple Vision Pro. Labeled as the first successful “spatial computer,” the gadget utilizes AI technology to process user queries and generate digital content.
The World Economic Forum in Switzerland launches AI Governance Alliance to address the issue of responsible AI, “uniting industry leaders, governments, and academic institutions.”
July
Anthropic further improves its generative AI solution with the release of Claude 2.
Meta releases LLaMA 2 together with Microsoft, making it free for research and commercial use.
At the annual HDC event, Huawei announces an upgrade of its smartphone virtual assistant, Celia, which now possesses generative AI capabilities. Backed by the company’s Pangu 3.0 LLM, the newest version of Celia is offered to the public within days as part of Huawei’s HarmonyOS 4.
The Chinese government publishes the finalized version of its Interim Measures for the Management of Generative Artificial Intelligence Services with the aim of forming a regulatory framework around generative AI.
August
South Korean internet giant, Naver, releases CLOVA X, a generative AI chatbot tailored for web browsing, which is based on the company’s 200-billion-parameter HyperCLOVA X language model.
The state of AI report from McKinsey & Company comes out, revealing that one-third of all surveyed organizations are “already regularly using generative AI in at least one function.”
The first mentions of Xiaomi’s MiLM-6B — a lightweight model for mobile devices — find their way into the press. Concurrently, the company’s virtual assistant, Xiao AI, gets a generative AI makeover.
September
Meta announces a beta release of Meta AI, a generative AI chatbot integrated within Meta’s product suite, that can generate text and photorealistic images.
Yandex introduces YandexGPT 2, featuring a 1.5-fold increase in training data and a 65% improvement in response quality, reportedly making this generative AI solution clever enough to qualify for entry into a Russian university.
OpenAI releases DALL-E 3, the latest version of its text-to-image visual generative AI solution, which is soon made available to Plus subscribers within the ChatGPT suite.
October
One of the planet’s largest AI conferences, World Summit AI, takes place in Amsterdam, the Netherlands. The event features prominent speakers from Amazon, Microsoft, Google, Nvidia, IBM, NASA, Mastercard, MIT, Berkeley, Harvard, the UN, the UK government, and the Council of Europe among numerous others.
Australia’s very first Generative AI Summit takes place in Sydney, with keynote appearances from representatives of the Australian government, ABC, Microsoft, Google, Optus, Telstra, and Sydney University.
November
The Bletchley Declaration is published by the UK government during the AI Safety Summit, focusing global efforts on responsible and ethical AI development.
Sam Altman, CEO of OpenAI, gets fired and then rehired by the board of directors in a matter of days amidst corporate turmoil. Microsoft gets a non-voting OpenAI board member seat as a result.
Anthropic releases Claude 2.1, increasing the AI bot’s processing power to 500 pages of typed text.
Amazon announces the preview of Amazon Q, a business-oriented generative AI assistant made for IT professionals and software developers.
December
Google announces the release of Gemini, its “by far the best we’ve got” multimodal language model, which is expected to be integrated into Bard shortly.
The European Parliament enacts the EU AI Act, initially proposed in 2021, aimed at ensuring the safe and ethical development and deployment of AI technology within the EU, for both commercial and private use.