On Monday 6th November 2023, OpenAI got to host its first developer conference dubbed 'OpenAI DevDay'. Live from San Francisco, a list of new features was announced, which I'll break down here in this article.
Since we discovered ChatGPT via OpenAI as it was open to the public, the Artificial Intelligence field has become disruptive in a year. More upgrades and features have been added in recent months and there were collaborations with Microsoft to power Bing Chat. There have also been other rival AI models like Bard by Google and the new Grok xAI by Elon Musk.
To be honest, I was not aware of the virtual event until I was notified by Brian a.k.a. Breezy and decided to check it out. If you didn't hear about it and you missed out, I got you covered.
Summary of the new stuff
- GPT-4 Turbo with 128K context - The next frontier revealed in OpenAI DevDay is the GPT-4 Turbo model. With a stunning 128K window, the AI Model is said to be "more capable and has knowledge of world events up to April 2023".It can fit an equivalent of more than 300 pages of text in a single prompt.
- Function calling updates - This one also has gotten a major upgrade. Improvements like the ability to call multiple functions in a single message and function calling accuracy were added.
Function calling lets you describe functions of your app or external APIs to models, and have the model intelligently choose to output a JSON object containing arguments to call those functions.
- Improved instruction following and JSON mode - GPT-4 Turbo has pushed the limit to perform better than their previous models on tasks that require careful following of instructions like generating specific formats. It introduces a new JSON mode, which ensures the model will respond with valid JSON.
- Reproducible outputs and log probabilities - The addition of the new seed parameter has the function of enabling reproducible outputs by making the model return consistent completions most of the time. OpenAI has been using this feature internally for its own tests.
- Updates to GPT-3.5 Turbo - GPT-4 is not the only model getting the "turbo" necessities. GPT-3.5 will also get the Turbo treatment that supports a 16K context window by default with 38% improvement in tasks like generation JSON, XML and YAML.
- Assistant API - This will help developers build agent-like experiences within their own applications. The API is designed for flexibility; use cases range from a natural language-based data analysis app.
- New modalities in the API - GPT-4 Turbo can accept images as inputs and developers can now integrate DALL-E 3. Developers can now generate human-quality speech from text via the text-to-speech API.
- Copyright Shield - OpenAI introduced Copyright Shield, a security feature that is committed to protecting its customers with built-in copyright safeguards in their systems. The company has also implemented safety measures to ensure responsible AI usage.
- Introduction to GPTs - GPTs are a new way for anyone to create a tailored version of ChatGPT to be more helpful in their daily life, at specific tasks, at work, or at home—and then share that creation with others.
