OpenAI DevDay Conference 2023: Key Highlights for Businesses and Organizations

OpenAI’s DevDay 2023 unveiled a wave of transformative tools, from GPT-4 Turbo and the Assistants API to customizable GPTs and copyright protection. These innovations are a major leap forward for developers building the next generation of AI-powered applications.

OpenAI’s first-ever developer conference, DevDay 2023, introduced a range of groundbreaking innovations that could revolutionize the tech industry, and showed its aim to shape the future of AI.

The latest models and improved features are designed to provide more control to developers, which will lead to the emergence of a new and unique ecosystem of applications.

Key new additions and improvements highlighted in OpenAI’s inaugural DevDay include:

GPT-4 Turbo comes with a 128K context window that can fit more than 300 pages of text in a single prompt and will be offered at a cheaper price. Turbo provides more control, especially in the context of model inputs and outputs.

Turbo updates the function calling feature by including the ability to call multiple functions in a single message. For example, users can “send one message requesting different actions, like opening the car window and turning off the A/C.”

It also provides improved instruction following (e.g., generating specific formats) and supports JSON mode, which is “useful for developers generating JSON in the Chat Completions API outside of function calling.”

Besides this, OpenAI will launch a feature enabling the return to log probabilities, which will be “useful for building features such as autocomplete in a search experience.” It will launch a new beta feature that “enables reproducible outputs, which is useful for use cases such as replaying requests for debugging, writing more comprehensive unit tests, and generally having a higher degree of control over the model behavior.”

The company will also release an updated GPT-3.5 Turbo that supports a 16K context window by default, improved instruction following, JSON mode, and parallel function calling.

This feature provides new capabilities such as Code Interpreter, Retrieval, and Function Calling. OpenAI clearly states that data and files provided to the OpenAI API are never used to train their models, and developers have the option to delete the data.

The new assistant API introduces persistent and infinitely long threads, allowing “developers to simply add each new message to an existing thread, hand off thread state management to OpenAI, and work around context window constraints.” The assistant API will enable developers to embed agents into solutions, building agent-like experiences within their own applications and providing customized solutions while sharing revenues with OpenAI.

New API modalities include:

Text-to-speech (TTS) which allows developers to generate human-quality speech from text.

DALL·E 3 incorporates built-in moderation to help developers protect their applications against misuse. Companies such as Snap and Coca-Cola have used DALL·E 3 to generate images and designs for their customers and campaigns.

GPT-4 Turbo with Vision will be able to “accept images as inputs, enabling use cases such as generating captions, analyzing real-world images in detail, and reading documents with figures.”

Custom models create an experimental access program for GPT-4 fine-tuning. OpenAI is launching a Custom Models program, providing an opportunity for selected organizations to work with OpenAI researchers to “train custom GPT-4 in their specific domain and provide organizations with exclusive access to their custom models.”

OpenAI also highlights that custom models will not be served or shared with other customers, or used to train other models. Moreover, the proprietary data provided to OpenAI to train such models will not be reused in another context.

OpenAI will now provide a Copyright Shield, like Microsoft and Google, which includes defending OpenAI’s customers and paying costs incurred in case of legal claims around copyright infringement. This will be applicable to the features of ChatGPT Enterprise and the OpenAI developer platform, but not to its free ChatGPT tool.

This implies that there will be an indemnification clause in the contract wherein OpenAI will defend a user against whom a copyright infringement claim (arising from the output generated by OpenAI’s tools) is made and pay for damages resulting from such claims. This may encourage more developers to choose OpenAI.

OpenAI will launch customizable GPTs to create a tailored version of ChatGPT for a specific purpose and share it with others. The GPT Store will be launched later this month and will feature creations by verified builders.

Chats with GPT will not be shared with builders, and OpenAI will provide enhanced privacy controls to users, such as choosing whether data can be sent to a third-party API if GPT uses one, providing the option to opt the user’s account out of model training, verifying builders’ identities, and preventing users from sharing harmful GPTs.

OpenAI will now provide cheaper APIs and higher rate limits. Besides this, OpenAI reiterates its enterprise privacy policies to assure that it abides by the principle of purpose limitation.

Recently, OpenAI LP (their for-profit Limited Partnership) has also revealed that the company is working on GPT-5 and plans to raise more funds from Microsoft.

The field of AI is evolving from an imperative technology to a declarative technology where users can simply state their desired outcome, and the tool provides it.

The field began with ChatGPT and ChatGPT API, which had user constraints, and has since progressed to encompass tools like ChatPDF (designed to answer questions about a document provided by the user), ChatGPT Wrappers (for instance, Jasper), and CanvaGPT (which allows users to build Canva posters). Now, customizable GPTs will encourage developers to build innovative and improved solutions with greater efficiency.

With these announcements at DevDay, OpenAI is providing advanced tools and resources for AI-driven innovation.

They are also fostering an OpenAI-centric ecosystem that is magnetizing individuals and businesses, making AI an integral part of their daily usage while adding further layers of abstraction.

https://openai.com/blog/new-models-and-developer-products-announced-at-devday

https://openai.com/blog/introducing-gpts

Ajay Mago

Ajay Mago is a former AmLaw 100 partner, with more than 16 years of experience in technology transactions, including at Jones Day, Mayer Brown, and Duane Morris. Mr. Mago advises on a wide array of sourcing, technology and other commercial transactions and helps companies navigate legal issues raised by data, emerging technologies, and digital transformation, both on the buyer and provider side. He routinely advises clients on IP strategy, management, and monetization arrangements, leveraging his technology background and business experience.