If the majority of announcements concern ChatGPT, which now has a tool belt, OpenAI intends to constitute a suite of products for all users of its LLMs. Here, we focus on the “star” ChatGPT, but LeMagIT publishes a second article dedicated to features targeting developers and researchers.
Firstly, there is the provision to all connected users of ChatGPT Search, a search engine tested in July, then formally presented last October. The tool makes it possible to anchor the responses of the major underlying language models (GPT-4o, GPT-4o mini) with results mainly coming from Microsoft Bing. The tool displays images and videos from YouTube in sessions with LLMs.
ChatGPT, promised to become the Gadget inspector of office automation
Last October, OpenAI also presented Canvas, a way to “collaborate” – or rather – to evolve a document iteratively using LLMs. It is now accessible to all users, like Artefacts in Anthropic's Claude.ai.
Thus, it is possible to edit a text or a portion of code. As part of text editing, Canvas can suggest changes, adjust content length, change the style according to the target audience, check syntax and grammar errors or even add emojis.
For developers, the tool provides shortcuts for reviewing code, commenting on it, adding logs, fixing bugs or even converting code to JavaScript, TypeScript, Python, Java, C++ or PHP.
In addition to Search and Canvas, OpenAI introduced Projects for paid ChatGPT Plus, Pro, and Team users. Projects must combine discussions, files, and personalized instructions in a single session. It's mostly a way to organize conversations into folders. OpenAI has also introduced a way to search for results in past conversations.
Clearly, it is a question of providing light generation capabilities augmented by research for small projects such as meeting preparation, learning or reviewing a subject. Later, OpenAI plans to connect this tool to Google Drive and Microsoft's OneDrive, as well as allowing the choice of the underlying LLM.
On paper, this is a competing feature to Google Cloud's Agentspace and NotebookLM, of which the provider recently presented a Pro and enterprise version. “Access [aux projets] will be extended to Enterprise and Edu users in January,” says OpenAI.
In its desire to bring together the greatest number of users who, possibly, would be more inclined to opt for a paid plan, the startup makes ChatGPT available from its website, a Windows, Android, iOS and macOS application.
Speaking of Apple, ChatGPT is integrated into its Apple Intelligence solution, on Mac, iPad and iPhone.
However, like the availability of the Sora and Sora Turbo video generation model, another announcement in the startup's calendar, users in the European Union do not yet have access to Apple Intelligence.
This is also the case for the “video” mode of Advanced Voice Mode, a feature launched in October 2024. When the advanced voice mode is sufficient to orally interrogate GPT-4o, this allows the tool to access a stream video through computer vision and screen sharing capability…except in the European Union and France.
Cloud giants flex their muscles
The team responsible for developing “Desktop” applications is currently developing “Work with Apps” based on the same capabilities, as well as “screen scraping” capabilities to enable LLM GPT-4o, o1, o1 mini and o1 pro to interact with software, starting with IDEs, including xcode, those from Jetbrains, Matlab or even Visual Code. In terms of office automation, the functionality allows you to interact with Apple Note, Quip and Notion. Work with Apps also includes advanced voice mode, but is currently only available for macOS. Windows support is coming “soon.”
OpenAI is clearly not alone in these different areas. In recent months, Amazon (Amazon Q), Google Cloud (Gemini for Workspace, Agentspace, Google AI Studio), Microsoft (Copilot), Anthropic (Claude.ai), Mistral AI (Le Chat) and others have tried to catch up their delay on the startup. The firm behind the Azure cloud and GCP seem to have found the formula to convert OpenAI customers.
Tech
