Discover more from Stemble - for the love of STEM!
Revolutionizing AI: Unveiling GPT-4 Turbo and other Game-Changing Developer Tools at DevDay
GPT-4 Turbo with 128K context and lower prices, the new Assistants API, GPT-4 Turbo with Vision, DALL·E 3 API, and more.
Dear Stemble Community,
We're excited to bring you the latest developments from the world of AI and technology in this week's edition of Stemble. Our focus today is on the groundbreaking announcements from DevDay, OpenAI's annual event. From the unveiling of the remarkable GPT-4 Turbo to new developer products and tools, there's something for everyone in the AI space.
GPT-4 Turbo with 128K context
We've been eagerly awaiting the next step in AI evolution, and OpenAI delivers with GPT-4 Turbo. This model comes equipped with an astounding 128K context window, allowing it to understand and respond to the equivalent of over 300 pages of text in one go. What's even more exciting is the 3x cheaper price for input tokens and 2x cheaper price for output tokens compared to GPT-4. For all developers, GPT-4 Turbo is now available in preview through the API, and a stable production-ready model is on the horizon.
Function Calling Updates
Function calling gets an upgrade with the ability to call multiple functions in a single message, simplifying the interaction with models and improving accuracy. GPT-4 Turbo is now more likely to return the right function parameters, making it an even more powerful tool for developers.
Improved Instruction Following and JSON Mode
GPT-4 Turbo excels in following instructions accurately, making it ideal for tasks that require specific formats. The introduction of JSON mode ensures syntactically correct JSON output, catering to developers working outside of function calling.
Reproducible Outputs and Log Probabilities
With the new seed parameter, developers can now achieve reproducible outputs, which is invaluable for debugging and unit testing. The introduction of log probabilities further enhances model control, enabling features like autocomplete in search experiences.
Updated GPT-3.5 Turbo
In addition to GPT-4 Turbo, a new version of GPT-3.5 Turbo is here, supporting a 16K context window by default. It offers improved instruction following, JSON mode, and parallel function calling. If you're currently using GPT-3.5 Turbo, you will be automatically upgraded to the new model on December 11, 2023. Older models will remain accessible until June 13, 2024.
Assistants API, Retrieval, and Code Interpreter
OpenAI is taking a significant step towards empowering developers to create agent-like experiences with the Assistants API. This new offering provides code interpretation, retrieval of external knowledge, and function calling capabilities to simplify the development of high-quality AI applications.
New Modalities in the API
GPT-4 Turbo with Vision
GPT-4 Turbo's vision capabilities are a game-changer, allowing it to process images and enable applications like image captions, real-world image analysis, and document reading. This opens up new possibilities for accessibility and image-related tasks.
DALL·E 3, a recent addition to the ChatGPT Plus and Enterprise users, can now be integrated into your apps and products through the Images API. It empowers you to generate images and designs programmatically, with built-in moderation to ensure responsible usage.
Generating human-quality speech from text is now possible with the Text-to-Speech API. With six preset voices and two model variants, this feature is optimized for real-time use cases and offers a high-quality option.
OpenAI is introducing an experimental access program for GPT-4 fine-tuning, allowing developers to tailor the model to their specific needs. Additionally, the Custom Models program provides organizations with the opportunity to work closely with OpenAI to create domain-specific models.
Lower Prices and Higher Rate Limits
Lower prices and higher rate limits make OpenAI's offerings even more accessible. For example, GPT-4 Turbo input tokens are now 3x cheaper than GPT-4, and GPT-3.5 Turbo input tokens are 3x cheaper than the previous 16K model. These adjustments aim to support the growth of your applications.
OpenAI is committed to protecting its customers and will now defend you, covering the costs, if you face legal claims around copyright infringement for generally available features of ChatGPT Enterprise and the developer platform.
Whisper v3 and Consistency Decoder
OpenAI is also rolling out Whisper large-v3 and open-sourcing the Consistency Decoder, which enhances speech recognition and image generation.
These updates from DevDay mark a significant leap in AI capabilities and accessibility, and we can't wait to see the incredible innovations that will arise from these new tools and models.
Thank you for being a part of Stemble's journey into the exciting world of STEM and AI.
Thanks for reading Stemble - for the love of STEM! Subscribe for free to receive new posts and support our work.