This will also interest you
[EN VIDÉO] Artificial Intelligence in space The next major astronomical projects will generate enormous quantities of data that humans will have to…
After many twists and turns, OpenAI has officially launched its Internet browsing function via ChatGPT. When made available to the public, the database operated by the famous chatbotchatbot generative was limited to only data available on the Web until September 2021. This restriction made it possible to ensure the stability of the AI and better control what it can generate. Last March, OpenAI began to open the floodgates of data on the Web. In the process, numerous misappropriations and abuses took place. So the publisher backtracked several times this summer, before reestablishing a connection to the Web via Bing at the end of September. But today, while it was still only an experimental version, the chatbot can now be used to navigate thanks to Bing’s Web indexing database. However, the offer is reserved for OpenAI Plus and Enterprise subscribers.
A chatbot that sees, hears and responds with information from the Web
What this changes is that in addition to being able to have answers based on “fresh” content from the Web, ChatGPTChatGPT is also mixed with generatorgenerator of Dall.E images in its third version. Images, as well as web content, can now be displayed directly in ChatGPT. Likewise, it is possible to search for answers using images by uploading a photo so that the chatbot can find all the information on it via the Web. With this update, the chatbot is therefore enriched with information from the live Web, images and also audio, because it is possible to communicate vocally with it. Is this time the right one? In any case, OpenAI seems sure of the reliability of its chatbot since this live web access function is activated by default for paying subscribers.
The American publisher Open AI has just announced that its conversational agent, ChatGPT, can now exceed “the murmur of knowledge” which until then blocked it in September 2021. It is now possible for it to have access to data after this date and in real time thanks to its integration into the search engine MicrosoftMicrosoft Bing.
Futura article with ETX Daily Up, published September 28, 2023
It’s been a long time coming: ChatGPT can now browse the entire Web and thus have access to the most recent information. Until now, the intelligent chatbot was limited to sources no later than September 2021. This new feature is available on ChatGPT Plus and should be available to everyone very soon.
Not a day goes by without a new announcement regarding ChatGPT. After integrating the DALL-E image generation tool into ChatGPT Plus and the promise for this autumnautumn new features to be able to speak with the conversational agent and analyze photos in addition to writing simple prompts, OpenAI breaks the barrier of time. Until then, the information made available to ChatGPT was limited to the end of 2021, excluding dedicated plugins for subscribers to its version premium.
Now it can take into account the most recent information available on the Web, thanks to Microsoft Bing. Note that ChatGPT systematically indicates the link to its sources, to be able to immediately verify the authenticity of the information provided.
To activate this feature, you must go to your account settings, in the “ Beta features » and activate “ Browse with Bing » (browse with Bing). You can then use it as you wish by precisely selecting this option (Browse avec Bing) in the GPT-4 options menu. Be careful though, it is not possible to combine the use of Bing and plugins. Browsing is available today for ChatGPT Plus users and will be rolled out to everyone. very soon ».
With this new functionality, ChatGPT joins certain competitors who already offer navigation across the entire Web, like Bing Chat, in the NavigatorNavigator Edge under Windows, or even Google Bard.
OpenAI releases several plugins for ChatGPT. One of them, currently in the Alpha test phase, allows him to retrieve information from the Web to enrich his answers. Is it a good idea ?
Article from Sylvain BigetSylvain Bigetpublished on March 23, 2023
One of the limitations of ChatGPT is that it is essentially stuck in these responses on dates, events and people whose information is before September 2021. Regular updates have kept the information up to date. later dates, but until now artificial intelligence did not know how to search for data on the Web on its own.
It’s now possible, thanks to plugins that OpenAI added yesterday. They allow it to access external sources and databases and even search for information directly on the Web. Currently, these options are only available in Alpha version. These plugins will only be accessible to a limited number of developers and users who have opted into the plan premium you chatbot.
The plugins will then be accessible to everyone after this testing phase. The plugin that precisely adds web search queries. If ChatGPT is deliberately restricted, it is precisely to better filter content in order to limit access to unreliable content. Knowing that, despite this control, ChatGPT can easily go wrong, direct access to search engines could well make the phenomenon worse.
We remember the unfortunate experience of FacebookFacebook with its BlenderBot 3.0 chatbot which quickly delved into conspiracy theories and anti-Semitism due to the viral content it had access to. Similarly, in 2021, OpenAI tried to experiment with connecting to the web with an AI called WebGPT. The system, which looked for answers on the Web, tended to draw on unreliable but popular sources.
It remains to be seen how ChatGPT will fare with this new feature. Among the other plugins, we also find a Python code interpretation function. It is able to support file upload to the chatbot and can also upload the generated file. This process makes it possible to carry out processing and convert files from one format to another. Furthermore, OpenAI explained that it wishes to develop its plugin policy for ChatGPT and allow companies, for example, to create them to meet their own needs.
rewrite this content and keep HTML tags