OpenAI DevDay Announced Major ChatGPT Updates! New Version Attracted Overwhelming Traffic, Causing Server Crash!

On November 6, OpenAI hosted its inaugural DevDay, where CEO Sam Altman announced several major updates. The highlights include:

• Released new GPT-4 Turbo model 
• More controllable outputs: enhanced function calls, JSON mode.
• Opened new APIs: DALLE-3, GPT-4 Turbo with vision, TTS, and Whisper V3.
• GPT-4 fine-tuning, custom models.
• GPTs: Custom ChatGPT.
• GPT Store coming soon.
• Assistants API: a closer experience to AI agents.

Release of GPT-4 Turbo Supporting 128k Context Window

GPT-4 Turbo's knowledge base has been updated up to April 2023, compared to previous versions which only included world knowledge up to September 2021. 
Additionally, GPT-4 Turbo supports a 128k context window, which OpenAI describes as "equivalent to over 300 pages of text in a single prompt," roughly the length of "Wuthering Heights" and "Harry Potter." 
After optimizing the model's performance, the cost for developers using Tokens has also been reduced.

Turbo also supports a new "JSON mode," ensuring model responses in efficient JSON, an open standard file format and data interchange format. OpenAI states this is very useful for web applications that transmit data. The new model also allows multiple function calls at the same time. 

Multimodal Capabilities

Multimodal capabilities were also a focus of the conference, for which OpenAI opened brand new APIs. 
GPT-4 Turbo integrates DALL·E 3, capable of accepting and processing image inputs (i.e., GPT-4 Turbo with vision), generating titles, analyzing real-world images, reading documents with charts, etc. 
Developers can access GPT-4 Turbo with vision through the API's gpt-4-vision-preview. OpenAI plans to provide visual support for the main GPT-4 Turbo model, with pricing depending on the size of the input image. 
In the text-to-speech domain, developers can now convert text into human-quality speech through the text-to-speech API. The new TTS model offers 6 preset voices and two model variants, tts-1 and tts-1-hd, with tts optimized for real-time use cases and tts-1-hd optimized for quality.

GPT Store Coming Soon

Sam Atlman announced at the Developer Day that ChatGPT will soon launch the GPT Store, undoubtedly the biggest highlight of the conference! 
GPTs allow users to create custom versions of ChatGPT without code, combining their own instructions, external knowledge, and capabilities. In other words, anyone can easily build their own GPT. 
OpenAI provided custom GPT examples at the event: Canva and ZapierAI, which can be generated through very simple interactions. 
From the demonstration, to create a GPT, OpenAI allows users to use a conversational AI model called GPT Builder, enabling users to build custom GPTs using natural language. 
Some custom GPTs are already available for trial by ChatGPT Plus and enterprise users. 
The GPT Store is expected to launch by the end of this month, allowing developers to share and publish their created GPTs. 

Release of Assistants API

The DevDay also saw the release of the Assistants API, the first step towards helping developers build agent-like experiences within their own applications. An assistant can be understood as an AI built for a specific purpose, with specific instructions, utilizing additional knowledge, and capable of calling models and tools to perform tasks. The new Assistants API offers new features such as a code interpreter, retrieval, and function calling, which can handle a lot of the heavy lifting you previously had to do yourself, allowing you to build high-quality AI apps.

• Code Interpreter: writes and runs Python code in a sandboxed execution environment, and can generate graphs and charts, and process files with diverse data and formatting. It allows your assistants to run code iteratively to solve challenging code and math problems, and more.
• Retrieval: augments the assistant with knowledge from outside our models, such as proprietary domain data, product information or documents provided by your users. This means you don’t need to compute and store embeddings for your documents, or implement chunking and search algorithms. The Assistants API optimizes what retrieval technique to use based on our experience building knowledge retrieval in ChatGPT.
• Function calling: enables assistants to invoke functions you define and incorporate the function response in their messages.

The beta version of Assistants API is available for trial at: https://platform.openai.com/playground?mode=assistant
That wraps up the key points from this press conference.
The DevDay was announced several months in advance, and in a brief forty-five minutes, CEO Sam Altman introduced many important updates and the future vision for the products. Participants were excited, and online users following ChatGPT expressed their surprise.

New ChatGPT Version Causes Server Crash Due to High Traffic

However, possibly due to the high level of attention drawn by the Developer Day, an overwhelming number of users went to experience the new features two days after the conference, leading to a crash of the ChatGPT servers. The outage occurred approximately between 21:30 and 23:30 on November 8. 

CEO Sam Altman apologized on X (formerly Twitter). 

After the ChatGPT server crash, the team quickly investigated and fixed the issue within about 2 hours.
The site reminds: Use ChatGPT responsibly to avoid further crashes or service instability! 
We will continue to closely monitor updates from the ChatGPT team and promptly convey the latest information to users. Thank you for your support and understanding of our website.
Source: https://openai.com/blog/new-models-and-developer-products-announced-at-devday