OpenAI continues to improve ChatGPT with features that move the chatbot away from its purely text-based roots. Paying subscribers will be able to upload documents that the AI will analyze to answer questions. Multimodal support is also included.
ChatGPT gets closer every day or almost to a particularly gifted connected assistant. After integration with Dall-E which allows it to generate images based on queries, as well as synthetic voice support, OpenAI’s chatbot is now able to analyze PDF documents and images.
ChatGPT alone chooses the model to respond
Until now, to extract or search for information in a PDF, you had to go through an extension (like AskYourPDF). Now, this support is native: simply drag or save the PDF into the bot, then ask it questions related to the document (processing may take a few moments). ChatGPT also supports images in the most common formats.
Read ChatGPT Plugins: here’s how to boost AI with over 800 extensions
Another new feature: it is no longer necessary to select an operating model, for example Bing, to obtain responses from the web. The bot is now able to choose alone the way in which it should respond to requests, with Bing, or even Dall-E or in Python code. It all depends on the nature of the user’s question.
These changes, some of which were part of ChatGPT’s Enterprise subscription, are now rolling out to ChatGPT Plus users, at $20 per month.