I’ve asked ChatGPT to search and summarize the highlights from the DevDay 2023 promoted by OpenAI earlier today, and I gotta say they line up very effectively with what I saw.

The topics listed below are the result of asking for a list, extending it a couple of times, filtering some items, deepening each topic and listing references, and finally edited to make it cohesive and better stylized.

GPT Store⌗

Perhaps one of the most impactful announcements was the introduction of the GPT Store, a marketplace where individuals can build, publish, and potentially monetize their own versions of GPT models. This platform will allow users to create GPTs for various purposes without requiring coding experience, and these can range from simple to complex applications​

GPT-4 Turbo⌗

GPT-4 Turbo represents a significant advancement in OpenAI’s language models. It features a large context window of 128,000 tokens, allowing it to process the equivalent of over 300 pages of text in a single prompt. This upgrade makes it ideal for handling extensive data sets or documents. Additionally, it’s more cost-effective, with input tokens priced 3x cheaper and output tokens 2x cheaper than its predecessor, GPT-4​

Multimodal Abilities⌗

GPT-4 Turbo now includes multimodal capabilities like vision and text-to-speech. It can process images and has DALL-E 3 integration for image generation. Additionally, new text-to-speech voices with varying quality options have been added​

Improved Function Calling⌗

OpenAI has enhanced the function calling capabilities of its models, allowing them to call multiple functions in a single prompt, thus streamlining developer interaction with AI models​

New Assistants API⌗

This API assists developers in building apps with persistent state and advanced AI capabilities, such as code execution, knowledge retrieval, and function calling. It is a significant step towards creating more sophisticated AI applications​

Voice Interaction for ChatGPT⌗

OpenAI has integrated voice interaction capabilities into ChatGPT, enabling users to converse with the AI using spoken dialogue. This enhancement is particularly useful for mobile applications and is being explored for smart hardware devices like speakers. It expands the accessibility of AI interactions to a wider audience.

Reproducible Outputs⌗

The introduction of a seed parameter enables models to produce consistent outputs, aiding in testing and debugging. Log probabilities for tokens have also been added as a feature​

Updated GPT-3.5 Turbo⌗

The new GPT-3.5 Turbo version comes with a 16K context window and boasts a 38% improvement in tasks like generating JSON, XML, and YAML. OpenAI has also announced various price reductions for this model​

Updated ChatGPT Interface⌗

OpenAI has revamped the user interface of ChatGPT, making it cleaner and more intuitive for users. This update likely includes improvements to the layout, design, and navigability, aimed at enhancing the overall user experience and making it easier to interact with the various AI tools available like DALL-E 3​​.

Updated ChatGPT Information⌗

ChatGPT’s knowledge has been updated to include information up to April 2023. Moreover, ChatGPT has enhanced search capabilities, allowing it to sift through PDFs and other documents to find relevant information. This means the AI model is not only more current but also more adept at handling a variety of data sources for information retrieval​1​.

OpenAI now offers custom model training services and has introduced the Copyright Shield to protect customers from copyright infringement claims.

Microsoft Partnership⌗

The collaboration with Microsoft, highlighted by CEO Satya Nadella’s appearance, indicates a deeper integration of OpenAI’s technologies with Microsoft’s cloud and hardware offerings. This partnership signifies a strategic commitment to the advancement of AI applications and infrastructure.

Startup Mentor Program⌗

A program aimed at assisting startups by offering access to specialized AI for business advice. It leverages AI’s ability to analyze vast amounts of data for insights and recommendations, aiding startups in decision-making processes.

Giveaway app live coding⌗

As a promotional event, OpenAI showcased a live demonstration of a giveaway app and distributed $500 in OpenAI credits to participants. This gesture showcased the power of function calling combined with the new modalities. Initially five attendees were selected to receive $500 in OpenAI credits. Later, in a display of generosity, OpenAI decided to award credits of equal value to all DevDay attendees​.

Diwali Celebration⌗

OpenAI celebrated Diwali by showcasing how AI, specifically DALL-E 3, can be used creatively in festivities. For instance, Coca-Cola employed DALL-E 3 to create themed Diwali cards, illustrating AI’s potential in marketing and design by blending technology with cultural traditions.

How DALL-E 3 sees the event⌗

OpenAI DevDay imagined by DALL-E 3 I asked DALL-E 3 to imagine the event and this is what it came up with.

I can assure you there were no holograms, nor it took place in an amphitheather with desks. I could ask for the seed to that image and iterate on those details, but let’s leave it for a future article…

Recap and video of the event⌗

If you wanna see the actual event, below is the full opening keynote:

Update: OpenAI has released their own recap of the event.