OpenAI has announced a suite of new developments including the GPT-4 Turbo model which boasts a 128K context window and lower pricing, the Assistants API for building AI apps, and multimodal capabilities such as vision and text-to-speech. GPT-4 Turbo, which can process the equivalent of over 300 text pages, is more efficient and knowledgeable about events up to April 2023. Enhancements in function calling allow for complex multi-action requests, and improved instruction following is now possible with JSON mode. Moreover, the updated GPT-3.5 Turbo now supports a 16K context window and has shown significant improvements in task performance.
OpenAI has introduced customizable versions of ChatGPT, known as GPTs, which allow users to tailor the AI to specific needs and tasks without coding. The GPT Store, to be launched later this month, will enable creators to share their GPTs and possibly monetize them based on usage. Privacy and safety have been emphasized, with users having control over their data and options to integrate GPTs with external APIs for real-world tasks. These advancements aim to further engage the community in AI tool development while ensuring the responsible use of such technologies.