OpenAI Teaches ChatGPT to Read PDFs and Pick Tools Without Being Asked
Nice, I can finally easily chat with my documents and papers through native ChatGPT interface (GPT-4) pic.twitter.com/MvsaypE7tS
— LDJ (@ldjconfirmed) October 29, 2023
OpenAI has introduced a set of upgrades to its ChatGPT Plus subscription service that bundles all of the generative AI chatbot’s tools into a single setting and opens up more options for document analysis. The improvements to OpenAI’s GPT-4 large language model mean users can upload PDFs and other kinds of documents while getting the AI to analyze data or produce images within a single conversation.
ChatGPT United
Really keeping tracks of multiple different steps well all in one response. pic.twitter.com/BEwuPfzjMI
— LDJ (@ldjconfirmed) October 29, 2023
ChatGPT’s unification of tools means that document analysis, web browsing, and DALL-E 3 image generation no longer require the user to toggle each tool on separately, forcing them to copy information between chat windows to combine functions. Users can upload a photo and ask for a variation of it. The AI can even add an object from a different image entirely to its next production. The tools are all together, so a single query can ask for information, charts, and visuals all at once and get a helpful response.
The update essentially renders redundant various one-off apps and plugins wrapping OpenAI’s models. For example, third-party PDF analysis tools are now unnecessary, given ChatGPT’s new built-in document features. The accessibility could even make the AI assistant a rival to Adobe Photoshop for basic image editing. That exclusion seems to apply even to the ChatGPT plugins OpenAI hyped up so much when they debuted. Arguably, any specialized third-party plugin is obsolete as long as ChatGPT can mimic the feature natively.
Is this AGI?
What you are seeing in this photo:
1. Vision on input image
2. Generated new image based on input
3. Vision on generated image to create report data
4. Save report data as CSV file for download.Possible now with GPT-4’s latest update. pic.twitter.com/3Ax72rAaGZ
— Bryan McAnulty (@BryanMcAnulty) October 29, 2023
The news is part of a slew of recent ChatGPT upgrades. OpenAI recently gave ChatGPT the ability to search the internet for answers to user queries and summarize website content and added multimodal options for ChatGPT that allow for verbal conversations and enable the AI to process images in addition to text.
The new features also come with an update to ChatGPT’s knowledge base, making it more current. The new cut-off date is September 2023. A more comprehensive platform opens up potential security concerns, however. Pages accessed could contain hidden malicious prompts to misuse ChatGPT’s programming capabilities for ill intents. The plan to address those issues will likely come when OpenAI officially announces the upgrades. That may be next week at OpenAI’s DevDay conference, where it is expected to unveil new tools and technologies. However, the bundled functionality shows OpenAI aggressively working to maximize ChatGPT productivity and capabilities. It provides a glimpse into the assistant’s future as a centralized gateway to OpenAI’s leading AI.
Follow @voicebotaiFollow @erichschwartz
OpenAI Turns ChatGPT into a Voice Assistant That Can See and Understand Images and Speech
OpenAI Teases DALL-E 3 Text-to-Image Generator Built Into ChatGPT