OpenAI's Developer Day Highlights

OpenAI has done and is doing wonders for the gargantuan realm of technology after only releasing its revolutionary large language model, ChatGPT, less than a year ago.

With a relatively quick adoption pace and high usage rate of 100 million weekly active users, in addition to 92% of current Fortune 500 companies utilizing the tech, OpenAI thrives on aiming even higher and achieving even more.

During the company’s inaugural developer conference, OpenAI DevDay, CEO Sam Altman presented new updates, betas, and releases, from overall lower prices for OpenAI subscriptions to assistants API, hoped to reap immense benefits to even more creative developers, increasing the current number of 2 million, and GPT wrapped startups worldwide.

Upgraded Models with Lower Prices

After launching GPT-4 in March of this year, the first new upgrade OpenAI unveiled was the GPT-4 Turbo. While users were given a context window of 8K up to 32K tokens with GPT-4, GPT-4 Turbo’s context length reaches 128K tokens, roughly equivalent to about 100K words or 300 pages of a book.

OpenAI has optimized the model by not only increasing its context length but also by improving its overall performance. The model has a JSON Mode toggle to ensure that it responds to valid JSON, useful for better API and function calling. OpenAI has also managed to update ChatGPT’s knowledge cut-off, from its previous access limited to September 2021, now up to April 2023.

The GPT-4 Turbo, which is only available via an API preview at the moment, comes in two modes, one focusing on text-analyzing and the other on understanding both text and images. The model will be available for the general subscriber public in the upcoming weeks.

In addition to the model’s upgraded features, OpenAI also announced that it is cheaper to run, with inputs costing $0.01 per 1K tokens and outputs costing $0.03 per 1K tokens, more or less three times cheaper than the company’s earlier subscription-based models.

Customizable GPTs in Store

In an effort to provide smarter, more customizable tools for developers and creators, OpenAI unveiled GPTs, which are various versions of ChatGPT tailored for specific purposes, available for developers to build.

Whether for fun or more productive use cases, these personalized GPTs can be built through natural language prompts and conversations with the chatbot without any coding knowledge necessary.

By accessing the GPT Builder, developers can let the builder know what it is they want to build with detailed instructions through the Create tab. The builder will then start suggesting details regarding the GPT, which, once confirmed, will become visible on a preview section placed on the exact same page right next to the builder, allowing developers to stay in one place throughout the entire process.

The GPT’s features are completely customizable. Developers can either ask the builder to make any necessary changes or visit the Configure tab, which includes the name, description, instructions, conversation starters, capabilities, and actions they can fill out themselves. The Configure tab also includes a Knowledge section, which allows developers to upload a file that provides extra knowledge for the GPT to be based on, in addition to the knowledge GPT-4 already has.

Developers have the option of publishing these GPTs privately with links for people to use, or in the case of enterprises, limited to employees in the company. However, these GPTs can also be released and shared publicly, through OpenAI’s new GPT Store, which the company plans to release later this month.

To foster a beneficial and safe ecosystem for users and creators, OpenAI will filter through GPTs according to regulations and policies before making them accessible worldwide. Creators with the most used and most useful GPTs will receive a share of OpenAI’s revenues.

According to a roundtable Q&A with Altman and OpenAI CTO Mira Murati, the company is still in the middle of figuring out how revenue sharing will work. At the moment, OpenAI plans to share a part of the ChatGPT subscription overall revenue, with a tier based on the number of active users and some category bonuses.

Assistants API in a Playground

Lastly, Altman unveiled the Assistants API, which developers can use to build “assistants” within their own applications. Each assistant can be built with specific instructions and a model for it to leverage. They are also equipped with knowledge bases so they can perform tasks based on what the applications are intended for.

The Assistants API is powered by tools such as the Code Interpreter, which can write and run Python codes as well as generate charts and process files, Retrieval for it to access knowledge from outside OpenAI’s models like enterprise documents, and a Function calling. In the Assistants playground, developers can explore its assistant’s capabilities as well as learn how to build an Assistant without writing any code.

Aside from assistants, OpenAI’s text-to-image model DALL-E 3 is now available through API, along with the company’s new text-to-speech Audio API available in six preset voices, which are Alloy, Echo, Fable, Onyx, Nova, and Shimer.

Security and Startup Benefits

With the goal of defending businesses utilizing GPT models from copyright claims, OpenAI announced a security program named Copyright Shield. The program provides legal fees for customers facing IP lawsuits based on several conditions.

OpenAI’s unveiling presents greater opportunities for developers and creators as well as GPT wrapper startups, as they can now utilize more powerful and upgraded capabilities. Whether by maximizing GPT-4 Turbo’s high performance and low cost or using Audio API’s features to create more personalized, human-like services, it is an exciting time for developers and startups alike.

Meme & AI-Generated Picture

Job Posting

• Block - Machine Learning Modeler, Commerce Fraud Risk - San Francisco, CA+ (Remote)

• CashApp - Machine Learning Modeler, Commerce Fraud Risk - San Francisco, CA (Hybrid)

• Grow Therapy - Data Analyst, Multiple Levels - New York City, NY+ (Remote/Hybrid)

• PrizePicks - Data Scientist, GameOps, ESports - US (Remote)

Promote your product/service to Digger Insights’ Community

Advertise with Digger Insights. Digger Insights’ Miners are professionals and business owners with diverse Industry backgrounds who are looking for interesting and helpful tools, products, services, jobs, events, apps, and books. Email us [email protected]

Your feedback would be greatly appreciated. Send it to [email protected]