OpenAI held its first DevDay conference on 6 November 2023 and introduced several new artificial intelligence capabilities. At Zeo, we believe the event could have a significant influence on the future of AI assistants and developer tools. This article reviews the main announcements and their potential impact.
GPT-4 Turbo: The Next Evolution of Language Models
One of the event's main announcements was GPT-4 Turbo. This GPT-4 Turbo release improved on the original GPT-4 in several important ways:
- Massive Content-Length: GPT-4 Turbo will be able to process content lengths of up to 128,000 tokens, 16 times longer than GPT-4, which has a capacity of 8,192 tokens. This increased capacity allows it to process longer conversations more accurately and reliably.
- Improved Functionality: Developers will now be able to run multiple functions from a single prompt and receive responses in JSON format. GPT-4 Turbo will also provide more reliable results when handling complex logic operations.
- Repeatable Outputs: For the first time, developers will be able to produce consistent model outputs using Data Seeding.
- More Current Information: GPT-4 Turbo has been updated with information about events up to April 2023 and will refresh its information every 6 months, helping it avoid outdated responses.
- Multi-Mode Models: GPT-4 Turbo will now support DALL-E rendering and state-of-the-art text-to-speech synthesis, enabling developers to build new types of applications.
- Cost Reductions: GPT-4 Turbo provides a price reduction of more than 2.75 times compared with GPT-4. Prompt tokens will now cost $0.01 per 1000 tokens, while completion tokens will cost $0.03 per 1000.
These improvements made GPT-4 Turbo the most advanced publicly available AI system at the time of the announcement. Its expanded capacity gives software developers more room to build agents that use deeper judgment and broader knowledge. The lower pricing could also make the technology economically viable for more applications.
Create Your Own AI Assistant Without Code
Perhaps the event's most important announcement was GPTs, customized versions of ChatGPT designed for specific domains. GPTs allow people to create AI assistants for their own needs without writing code.
Users will be able to create GPTs with GPT Builder by:
- Providing Clear Guidelines - Assistants will be trained on how to respond in different scenarios.
- Including Relevant Data - Documents, spreadsheets, and APIs can be added to support integration.
- Enabling Actions - API calls and integrations can allow the assistant to perform different tasks.
You can also explore Zeo’s "Jarvis The SEO Geek," powered by ChatGPT. Jarvis draws on Zeo's #SEO knowledge through technical documents and case studies, and it has additional capabilities. It integrates APIs such as Google PageSpeed Insights and Ahrefs to provide more comprehensive website analysis. Jarvis is now live.
GPTs bring together three key elements of an AI agent: instructions, knowledge, and skills. This could allow everyday users to turn ChatGPT into an assistant tailored to their specific needs.
OpenAI also introduced a GPT marketplace where creators can publish their custom assistants. It can be thought of as a type of App Store. OpenAI said the best GPTs would receive revenue sharing, creating a new ecosystem around these tools. This change could make personalized AI assistants available to a much wider audience.
New Developer Building Blocks for Next-Generation Assistants
OpenAI also announced new APIs designed to make advanced AI assistants easier for developers to build:
- Persistent Topics: Automatically maintain a conversation state between messages.
- Information Storage/Recall: Add your own data sources and recall them in context.
- Code Interpreter: Write code through prompts and establish logic immediately.
- Streamlined Functions: Call multiple functions from one prompt and receive structured output.
These developer tools marked an important step toward wider access to the latest AI technology. They could let developers spend less time building underlying infrastructure and more time creating useful AI-powered experiences.
Secure and Iterative Development of AI Capabilities
During the keynote, OpenAI emphasized its careful approach to developing AI systems. The company was initially focusing on limited applications such as GPTs.
As AI capabilities grow, OpenAI is focusing on systematic testing and user feedback to address potential risks responsibly. Its stated goal is to assess and mitigate hazards while giving people access to useful AI capabilities. Prioritizing safety could help the company provide better assistance as these systems develop.
Strategic Partnership with Microsoft to Industrialise Artificial Intelligence
OpenAI and Microsoft have strengthened their partnership to provide the infrastructure needed to power and scale AI systems worldwide.
Microsoft is committed to providing Azure capabilities optimized for AI workloads. It is also integrating OpenAI APIs into its developer products, giving software engineers access to advanced AI capabilities.
The two companies are working to make AI more accessible to businesses around the world by industrializing the technology and providing enterprise APIs.
Conclusion: A Milestone for Artificial Intelligence
OpenAI's DevDay conference marked an important point in the development of more capable, affordable, and widely available AI assistants. New products such as GPTs and the developer APIs could make customized AI accessible to many more people.
Enterprises and developers will be able to use OpenAI's AI technology to create a new generation of intelligent agents and experiences. I expect these innovations to support useful applications across industries, while the company's iterative approach offers one model for developing AI more safely.
Overall, the DevDay announcements represented a substantial advance in AI systems designed to understand, reason, and assist. They also laid some of the groundwork for AI tools to become more common in homes and workplaces.
This content was created by Ahmet Can Kavak, Marketing Manager at Zeo.
