OpenAI releases GPT-4 — available via ChatGPT
OpenAI has just announced the release of GPT-4, their most advanced language model to date. The new system promises to produce safer and more useful responses, as well as solving difficult problems with greater accuracy thanks to its broader general knowledge and problem-solving abilities.
One of the most exciting features of GPT-4 is its increased creativity and collaboration. The model can now generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style. In addition, GPT-4 can now accept images as inputs and generate captions, classifications, and analyses.
Another major improvement is GPT-4's ability to handle over 25,000 words of text, allowing for use cases like long-form content creation, extended conversations, and document search and analysis.
OpenAI spent six months making the GPT-4 safer and more aligned, resulting in an 82% decrease in the likelihood of responding to requests for disallowed content and a 40% increase in producing factual responses compared to GPT-3.5 on internal evaluations.
OpenAI achieved this by incorporating more human feedback, including feedback submitted by ChatGPT users, to improve GPT-4’s behavior. They also worked with over 50 experts for early feedback in domains including AI safety and security. Furthermore, OpenAI has applied lessons from real-world use of their previous models into GPT-4’s safety research and monitoring system.
GPT-4's advanced reasoning and instruction-following capabilities have also been leveraged to expedite OpenAI's safety work. The model was used to create training data for model fine-tuning and iterate on classifiers across training, evaluations, and monitoring.
In addition, OpenAI notes that GPT-4 surpasses ChatGPT in its advanced reasoning capabilities and outperforms ChatGPT by scoring in higher approximate percentiles among test-takers.
As with ChatGPT, OpenAI will be updating and improving GPT-4 at a regular cadence as more people use it. GPT-4's advanced features and safety improvements make it an exciting addition to the world of natural language processing, with potential use cases in a wide range of industries, from healthcare to finance to entertainment.
Where GPT-4 is used:
Be My Eyes
Be My Eyes uses GPT-4 to transform visual accessibility.
Stripe leverages GPT-4 to streamline user experience and combat fraud.
Morgan Stanley wealth management deploys GPT-4 to organize its vast knowledge base.
Khan Academy explores the potential for GPT-4 in a limited pilot program.
Government of Iceland
How Iceland is using GPT-4 to preserve its language.
- New jailbreak for ChatGPT — March 8, 2023
- A guide to using ChatGPT for writing fanfiction about Voldemort's regime
- Now you can chat with your PDF with a new tool based on ChatGPT