What Went Down During OpenAI's DevDay?

OpenAI just announced a series of new services and products, including notable upgrades in its first-ever developer conference, DevDay.

We're rolling out new features and improvements that developers have been asking for:

1. Our new model GPT-4 Turbo supports 128K context and has fresher knowledge than GPT-4. Its input and output tokens are respectively 3× and 2× less expensive than GPT-4. It’s available now to…

— OpenAI (@OpenAI) November 6, 2023

GPT-4 Turbo With 128K Content

Following its initial release in March, the ChatGPT's most powerful model to date GPT-4 is finally getting a Turbo upgrade, with knowledge of events that can date up to April 2023.

The new AI tool will also come with a 128k context window that allows it to fit over 300 pages of text in a single prompt and will be three times cheaper for input tokens and two times cheaper for output tokens compared to its price upon release. 

A new experimental access program for fine-tuning GPT-4 will also become available, as the company admits to the model requiring "more work to achieve meaningful improvements." 

Additionally, OpenAI updated its GPT-3.5 Turbo model to support a 16k context window in its default mode, allowing instructional improvements, a JSON mode, and parallel function calling.

Previously, OpenAI made the Turbo version available to developers for fine-tuning. 

New Assistants API 

OpenAI has also launched the Assistants API, offering developers the tools to integrate specialized AI assistants into their applications.

These assistants are tailor-made AI entities with specific instructions, expanded knowledge, and the capability to do tasks using various models and tools.

The API introduces features several new features that streamline the development process for high-quality AI applications:

• Code Interpreter, which creates and executes Python code, makes graphs, works with various data and helps solve complex coding and math problems, allowing assistants to run code step-by-step
• Retrieval, which empowers the assistant with external knowledge, eliminating the need for developers to handle complex document processing. The API selects the best retrieval method based on our expertise in knowledge retrieval from ChatGPT
• Function Calling, which lets assistants use functions developers create and include their responses in their messages

OpenAI guarantees that data shared with the API won't train the model and can be deleted by developers as necessary.

Additionally, the company will be making GPT-4 Turbo with Vision, DALL-E 3, and text-to-speech available to developers.

ChatGPT Celebrates 100M Weekly Users

OpenAI CEO Sam Altman has confirmed the company's newest milestone: Achieving 100 million weekly ChatGPT users.

Alongside the momentous feat, Altman also revealed that OpenAI has garnered over two million developers, with 92% of them coming from Fortune 500 companies.

Previously, the chatbot broke records for becoming one of the fastest-growing internet apps of all time, accumulating over 100 million users in its first two months. 

OpenAI partners have also shared in the company's successes, with Microsoft's Bing Chat, powered by GPT-4, surpassing over 100 million users in March.