OpenAI DevDay 2023: 5 Key Takeaways from the conference

On Monday, OpenAI hosted its inaugural developer conference in San Francisco, DevDay. The one-day event was jam-packed with keynotes, live demos, and deep dive sessions. It was a landmark moment for developers, and I had a front-row seat to all the excitement!

OpenAI also announced new AI capabilities and enhancements that promise to change the game in terms of creating, deploying, and scaling AI applications:

• With the debut of GPT-4 Turbo, developers were given a glimpse into high-caliber AI efficiency. 
• The Assistants API emerged as a tool for those building more sophisticated and interactive AI-driven applications, while custom GPTs and the GPT Store introduced a collaborative AI marketplace with revenue sharing. 
• The introduction of Custom Models offers fine-grain control over AI behavior and a new service called Copyright Shield that provides robust privacy measures reinforcing OpenAI's commitment to secure, ethical AI practices. 

For developers wanting to join the AI innovation wave, these takeaways signal an easy entry point and a huge leap forward. Below are my top takeaways from OpenAI DevDay, which I feel encapsulate the event's most impactful announcements, highlight significant strides in AI technology, and offer developers new tools and capabilities to enhance their applications while emphasizing the importance of user trust and safety in the expanding landscape of AI.

1. GPT-4 Turbo
2. Assistants API
3. Custom GPTs and the GPT Store
4. Custom Models 
5. Copyright Shield

The new announcements around GPT-4 Turbo from OpenAI DevDay included introducing a more capable model that supports a 128K token context window and is priced more affordably than GPT-4. GPT-4 Turbo is designed to understand and generate responses based on a much larger context, effectively handling the equivalent of over 300 pages of text in a single prompt. Additionally, it's been optimized for performance, resulting in a 3x lower price for input tokens and a 2x lower price for output tokens compared to the standard GPT-4. While this is all great, what excites me the most is that GPT-4 has knowledge of world events up to April 2023!

OpenAI also plans to roll out new features for GPT-4 Turbo, including function calling improvements and a new JSON mode for developers. This means developers can now build applications that maintain continuity over lengthy interactions, ideal for detailed customer service, long educational dialogues, or in-depth content creation. The performance optimization leading to lower costs makes GPT-4 Turbo a more economical choice for developers and businesses.

The Assistants API is a new feature that allows developers to build more complex, agent-like AI applications. This API equips developers with the tools to build AI assistants that not only understand complex commands but can also leverage an array of models and tools to carry out tasks.

From executing code on the fly using Code Interpreter to using Retrieval to enhance AI responses using external data stores, this API enables the creation of AI with more autonomous, intelligent behaviors. Its support for persistent conversation threads also provides a new level of continuity in AI interactions, ideal for complex customer service scenarios or interactive educational platforms. This API will usher in the next generation of interactive AI!

Custom GPTs are personalized versions of ChatGPT that developers or organizations can tailor for specific tasks or roles. What is really cool is that GPTs can be created without coding, using natural language (i.e., simple instructions) to customize their capabilities.

The GPT Store, which is set to roll out, will be a marketplace where these custom GPTs can be shared and monetized. Think of it as an app store for AI, where developers can publish their GPTs, and users can discover a variety of purpose-built AI tools. What’s really cool is that developers can potentially earn revenue based on user engagement.

The store aims to foster a community of AI builders and users, promoting innovation and collaboration in the AI space. I had an opportunity to try out GPT Builder—a tool for building GPTs—and could customize ChatGPT with just a few short instructions in under five minutes, with no coding involved!

Custom models are specialized AI models tailored to the specific needs of an organization, beneficial for domains with large, proprietary datasets. These models are created in collaboration with OpenAI researchers and allow for deep customization, including the option to adjust every step of the model training process. This approach means businesses aren't just applying AI—they're integrating it with their unique workflows and data ecosystems, giving them a competitive edge.

Organizations that opt for custom models will have exclusive access to them, ensuring that proprietary data used for training is not shared or reused elsewhere. This program is designed to meet the needs of enterprises requiring AI capabilities that are beyond the scope of standard fine-tuning.

Custom models do appear to be a costly option and will probably only be accessible to organizations with hefty AI budgets—interested organizations can apply here.

Copyright Shield is a new feature designed to offer legal protection to users of the OpenAI platform. It indicates OpenAI's commitment to defend and cover the costs for its customers if they face legal claims of copyright infringement related to the use of OpenAI's generally available features of ChatGPT Enterprise and the developer platform. This initiative reflects an effort to address the complex legalities associated with the creation of content by AI and to provide a layer of security for developers and businesses using the platform. 

This proactive step by OpenAI with Copyright Shield provides a safety net that could significantly lower the barrier to entry for developers and businesses interested in AI. It addresses a critical aspect of content generation in the age of AI: the legal responsibility associated with AI-created content.

With this measure, OpenAI is not just offering a technological tool but also ensuring peace of mind, allowing innovators to focus on creativity and development without the looming concern of legal repercussions from inadvertently generated content.

As we reflect on the groundbreaking advancements revealed at OpenAI DevDay, it's clear that the path ahead for developers is both exciting and promising. The event has opened new paths to innovation with the unveiling of GPT-4 Turbo, Assistants API, Custom GPTs, and the GPT Store, each providing an opportunity to revolutionize our digital experience.

With the added security of Copyright Shield, developers can now navigate this new landscape with confidence and creativity. At Pluralsight, we're excited to see how these innovations will shape the future of technology, and we remain committed to empowering developers with the skills needed to thrive in this bold new era of AI.

If you’re just beginning with Generative AI and ChatGPT, learning prompt engineering is a great starting point.  Take a look at my course "Prompt Engineering for Improved Performance" to master advanced techniques in prompt engineering! Pluralsight also offers a range of beginner, intermediate, and expert AI and ML courses for you to check out.