The new version of the company's AI model can understand and respond to image inputs.
OpenAI, the development company behind the image generator DALL-E and chatbot ChatGPT, has released GPT-4, a new version of their AI model. Trained on publically available data, up to September 2021, and human feedback, GPT-4 is a multimodal model, meaning it is capable of understanding and responding to both text and image inputs and can produce text outputs based on them, unlike its predecessor GPT-3.5, which could only understand the text.
According to OpenAI, they spent half a year iteratively aligning GPT-4 using lessons from the adversarial testing program and ChatGPT, resulting in the "best-ever results on factuality, steerability, and refusing to go outside of guardrails."
The highlight of the new AI model is, without a doubt, its ability to understand images and give a response based on the text input provided alongside the image. GPT-4 demonstrates comparable abilities across various domains, encompassing documents that comprise text, photographs, diagrams, or screenshots, the team noted. Additionally, it can be enhanced with test-time approaches that have been devised for text-only language models, such as few-shot and chain-of-thought prompting.
"A year ago, we trained GPT-3.5 as the first "test run" of the system," comments OpenAI. "We found and fixed some bugs and improved our theoretical foundations. As a result, our GPT-4 training run was unprecedentedly stable, becoming our first large model whose training performance we were able to accurately predict ahead of time. As we continue to focus on reliable scaling, we aim to hone our methodology to help us predict and prepare for future capabilities increasingly far in advance – something we view as critical for safety."
At the moment, GPT-4 is available via ChatGPT Plus, a pay-to-use version of ChatGPT, or by joining the GPT-4 API waitlist. You can learn more about the new model by checking out OpenAI's official announcement.
Also, don't forget to join our 80 Level Talent platform and our Telegram channel, follow us on Instagram and Twitter, where we share breakdowns, the latest news, awesome artworks, and more.