2023

February

6 events

Editors' summary

Rivals unveil their own AI, ChatGPT starts charging

On 6 February, Google announced Bard. Microsoft unveiled its AI-powered Bing the next day; the timing looked designed to get out in front of it. But Bard's demo contained a wrong answer about space telescopes. Once that was pointed out, Alphabet's stock fell, and the view that Google was rattled was reinforced instead.

The same month, OpenAI launched ChatGPT Plus at $20 a month and began earning revenue. Google, meanwhile, was putting money into Anthropic. At the end of the month Meta released LLaMA for research use only. It would change what came next.

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

Record6 events
  1. 1
    ProductOpenAI

    ChatGPT Plus subscription plan announced

    OpenAI announced ChatGPT Plus, a paid subscription plan for ChatGPT priced at $20 per month, offering general access even during peak times, faster response times, and priority access to new features, initially rolling out in the United States. Free access continued, but two months after launch OpenAI had a paid tier in place.

  2. PolicyGetty Images / Stability AI

    Getty Images sues Stability AI over 12 million images

    Getty Images sued Stability AI in the US District Court for Delaware, alleging that it had copied and processed 12 million Getty images together with their captions and metadata, without a licence, to train Stable Diffusion. Getty said the company had "chosen to ignore viable licensing options and long-standing legal protections in pursuit of their standalone commercial interests." Where the artists' class action the previous month was about individual works, this came from a business built on licensing images, and its argument was that a licensing route existed and was not taken. The dispute over training data now ran across rights holders of every size.

  3. 3
    BusinessReportedAnthropic / Google / Dario Amodei

    Google reported to have invested about $300 million in Anthropic

    The Financial Times reported that Google had invested about $300 million in Anthropic for a stake of roughly 10 percent, at a valuation of around $5 billion, with the money reportedly tied to a commitment for Anthropic to buy compute from Google Cloud. Coming days after Microsoft announced its large follow-on investment in OpenAI, the deal was read as confirmation that major cloud providers were locking in the leading AI research labs. Anthropic had not yet released a public product at the time.

  4. 6
    ProductGoogle / Sundar Pichai

    Google announces Bard, then a wrong answer wipes out $100 billion

    Sundar Pichai announced Bard, Google's conversational AI service, in a blog post. Built on a lightweight version of LaMDA, it would go first to "trusted testers" before a public release in the following weeks. The announcement came about two months after ChatGPT launched and less than two weeks after Google was reported to have declared an internal "code red." But in the promotional material, Bard gave a wrong answer about which telescope first photographed a planet outside our solar system, crediting the James Webb Space Telescope when earlier images had come from another instrument. Reuters spotted the error, and combined with a Paris AI event on February 8 that failed to impress, Alphabet's shares fell more than 7% that day, erasing roughly $100 billion in market value. It was widely reported at the time as a rushed announcement that had backfired.

  5. 7
    ProductMicrosoft / OpenAI / Satya Nadella

    Microsoft unveils AI-powered new Bing and Edge

    At an event at its headquarters, Microsoft unveiled a new AI-powered Bing search engine and Edge browser running on a next-generation OpenAI model, described as more powerful than ChatGPT and customized for search, positioning them as a "copilot for the web." The launch sharpened Microsoft's challenge to Google in search and signaled the start of an all-out generative AI race among big tech companies.

  6. 24
    ModelMeta

    Meta releases LLaMA to researchers

    Meta released LLaMA in four sizes — 7, 13, 33, and 65 billion parameters — distributing the weights only to researchers and government applicants who passed an approval process, under a non-commercial research licence. The paper reported that the 13B model beat GPT-3 (175B) on most benchmarks and that the 65B was competitive with DeepMind's Chinchilla-70B and Google's PaLM-540B: train a smaller model for longer, and it catches up with far larger ones. Handing out weights instead of opening an API was a limited, research-only measure at this point, but it put a frontier-class model that ran on modest hardware outside the companies that built it.