During yesterday’s DevDay keynote, OpenAI announced a series of new additions and updates to their platform, born to impress!
OpenAI introduced GPT-4 Turbo, an extremely advanced model knowledgeable about global events up to April 2023. Capable of processing 128,000 input tokens, equivalent to over 300 pages of text per prompt, this model is ready to write about tomorrow’s chapter. Interestingly, the model is not only smarter but also more cost-effective with significantly reduced API prices for input and output tokens.
With the new updates, it is now possible to call all functions in a single chat. Additionally, the precision of the functions has improved, meaning you can access the correct functions as easily as throwing a lasso in a rodeo – yeehaw! Or at least, “Browse with Bing” should figure out when to start and fetch content from a URL in your prompt on its own.
Not only that, but both GPT-4 Turbo and the new improved and much faster ChatGPT-3.5 have enhanced the ability to follow instructions more closely. The new JSON mode ensures that the model responds with valid JSON output – it’s like checking the mail and only finding stickers and candy instead of bills!
OpenAI has introduced a ‘seed’ parameter in API calls, ensuring that the AI’s responses are much more consistent most of the time. The intention is to create a more coherent experience. Moreover, they are working on introducing a feature that returns a probability-divided list of output tokens, a crucial component for creating razor-sharp autocomplete functions in, for example, searches.
An updated version of GPT-3.5 Turbo is ready, and just like GTP-4-Turbo, it comes with improved instruction compliance, JSON mode, and parallel function calling. OpenAI has also introduced the Assistants API, allowing developers to build agent-like experiences in their applications. Assistants can call models and tools to perform tasks – they’re like little AI butlers!
Everyone will be able to build AI agents, and even put them up for sale in the OpenAI store, where others can purchase access and use these agents to solve tasks.
GPT-4 Turbo can now accept images as input in the Chat Completions API, DALL·E 3 can now be integrated into applications via the Images API, and OpenAI has introduced a text-to-speech API that can produce human-like voice from text. It is truly a wonderful time for developers!
Several other exciting updates have also been announced. From GPT-4 fine-tuning, tailor-made models for organizations, and reduced prices for certain models, to higher speed limits and the introduction of Copyright Shield.
The highlights from OpenAI’s DevDay event clearly signal their dedication to improving the platform and offering new features and products to both developers and regular users. With their focus on innovation, OpenAI has laid a solid foundation for new and exciting applications built on their technology. So, prepare for a new era of AI!”
