It continues to move on the side of Generative AI and will continue to do so over the holidays. OpenAI opens the final API of GPT-4 and launches a code interpreter in ChatGPT.
There is definitely always something happening on the Generative AI side. And especially on the OpenAI side. With good and bad news.
Upcoming trial and stagnant hearing
On the bad news side, OpenAI noted that authors Sarah Silverman, Richard Kadrey, and Christopher Golden sued them (and Meta) for copyright infringement after discovering in their pre-release that their fonts had been used to power AI engines. The plaintiffs allege that leaked information about the company’s AI activities shows that their work was used without permission. They submit to OpenAI that the complainants’ working summaries prepared by ChatGPT clearly indicated that the conversational AI was trained on their protected content. A case to pursue as it could lead to many more lawsuits later.
Another bad news: ChatGPT audience would have dropped by 10% between May and June. There is no doubt that the “curiosity” effect is wearing off. And that the competition from Bing Chat and Bard can be felt a bit. However, SimilarWeb’s audience measurement doesn’t take into account what really matters to OpenAI: subscribers to the paid version “ChatGPT Plus” (which are growing in number) and calls to its APIs from other applications. and IA (again, these are calls). rewarding for OpenAI).
SimilarWeb also notes that Bing Chat has a fairly stable audience at almost 1,250,000,000 visits per month, and significantly more than two other competitors, CharacterAI and Google BARD, which would have under 250,000 monthly visits.
GPT-4 going mainstream
The good news is that OpenAI is continuing on its way. The company has just opened its first office outside of the US (in London) and has just assembled a new team of its most brilliant researchers and engineers to ensure it doesn’t harm tomorrow’s superintelligences (AGIs or artificial general intelligence) or even humanity destroy: “In the next 4 years we will use 20% of the computing power available to us to solve the problem of superintelligence alignment.” Our main goal in basic research is our new superintelligence alignment team, but achieving this is part of our mission is essential and we expect many teams to contribute, from the development of new methods to their deployment,” explains the editor.
Another piece of good news: the GPT 4 API is now “Generally Available”. Currently, only OpenAI users who have already made payments have access to it. But in the long run, everyone will have access to it for a fee. Previously, access to the API was restricted to developers selected by OpenAI. Please note, however, that despite this “general availability” there are restrictions. Thus, access to the image analysis is restricted to developers selected by OpenAI. Likewise, the discussion window remains limited to 8,000 words. Finally, the API still seems limited to 25 messages every 3 hours. Limits that are likely to change quickly, especially since the publisher has announced that the old versions of GPT will soon become unavailable and will be replaced by “GPT 3.5 Turbo”.
OpenAI has also confirmed that the APIs DALL-E 2 (image generation) and Whisper (speech transcription) will also be moving to “general availability”.
A code interpreter in ChatGPT Plus
Finally, the last piece of good news for those who code with ChatGPT: A “Code Interpreter” plugin is now available for ChatGPT Plus users. This “code interpreter” allows ChatGPT to run Python code, not just generate it. This usually allows ChatGPT to create charts, maps, data visualizations and graphs, format data instantly, clean datasets, analyze music playlists, edit video files, create interactive HTML files, load an image to then to save, etc. The interpreter opens up a wide range of functions and allows the AI access to files (both reading and writing), making it a powerful tool for visualization, analysis and manipulation. Data.