Readers like you help support MUO. If you make a purchase through links on our site, we may receive an affiliate commission. Continue reading.
OpenAI has released GPT-4, an update to the artificial intelligence model powering its impressive and hugely popular generative AI ChatGPT.
GPT-4 brings huge upgrades to the model, which can now accept and process data in both text and images, and then respond appropriately to user input.
GPT-4 is a huge leap forward, with the ability to handle more words, formats, and data than GPT-3.5, and it could change more than we know.
OpenAI unveils improved GPT 4 model
on March 14, 2023, OpenAI has announced GPT-4, its updated artificial intelligence model. GPT-4 is the culmination of a two-year process in which OpenAI rebuilt its “entire deep learning stack”, fixed ChatGPT bugs and AI alignment issues, and built on the resounding success of ChatGPT (which uses GPT-3.5).
A year ago, we trained GPT-3.5 as the first “test run” of the system. We found and fixed some bugs and improved our theoretical foundations. As a result, our GPT-4 training run was unprecedentedly stable (at least for us!), becoming our first major model whose training performance we could accurately predict in advance.
The biggest change for most users is that GPT-4 can now accept a prompt that contains images and text instead of just text. Additionally, the new model can generate contextual responses based on the image you upload, creating human-like feedback even when multiple images are displayed.
OpenAI hasn’t mentioned video processing yet, but imagine that video processing and analytics is on the horizon – or closer.
OpenAI presents GPT-4 with impressive examples
OpenAI also showcased GPT-4’s capabilities in a live stream, which you can watch below.
The developer livestream showed how GPT-4 could first be used to build a new Discord bot from scratch, which could then build a working website from a roughly drawn image with a few symbolic words. Was it easy? Yes. Did it work within seconds? Absolutely.
The live stream mainly focused on how developers can use GPT-4, but was also an impressive demonstration of GPT-4’s capabilities that will continue to improve, especially compared to what you can do with ChatGPT.
GPT-4 is now available
GPT-4 is already available for Premium ChatGPT users. The new GPT 4 model went live with the announcement and users are already deploying the updated model.
However, OpenAI emphasized that while GPT-4 is a significant upgrade, it still has “similar limitations to previous GPT models”. GPT-4 will still lie, make untruths, make mistakes in reasoning and “hallucinate facts”. However, these moments should be significantly reduced, making GPT-4 outputs much more trustworthy and accurate than GPT-3.5.
It’s a remarkable advance for generative AI, that’s for sure.
This article was previously published on Source link
