AI Weekly Digest #27: OpenAI Major Updates

Hello, tech enthusiasts! This is Wassim Jouini and Welcome to my AI newsletter, where I bring you the latest advancements in Artificial Intelligence without the unnecessary hype.

You can find me on LinkedIn, Twitter and Medium! Let’s connect!

Now let's dive into this week's news and explore the practical applications of AI across various sectors.

Main Headlines

Obviously today, we'll cover OpenAI's Developer Conference Announcements. There's a lot to unpack, but I'll keep it brief and focus on what you should remember and how it will impact your business.

#1 GPT-4-turbo, Faster, Cheaper and Multimodal model!

Since March 2023, and OpenAI's GPT-4 Vision demo showcasing its ability to analyze images, generate code from sketches, interpret memes, and so on, we've been eagerly awaiting the opportunity to leverage such capabilities in our products and processes.

Yet, the size of the model, its cost, as well as the overall GPU shortage, forced OpenAI to delay the release of the multimodal model in favor of a more business-oriented version. A version that is cheaper and faster, even if it often means having a slightly less capable model.(read: “OpenAI Models’ Drift, Fast & Stupid? — New Insights and Consequences”).

That's how we got GPT-3.5-Turbo, which is much faster and much cheaper than GPT-3.5 (thanks to a distilled 20B parameter model instead of the original 175B parameter model). And that's probably how we're getting the new GPT-4-Turbo version!

Main GPT-4-Turbo announcement :

• Faster: almost as fast as gpt3.5-turbo!

• Cheaper API: 1ct / 1k token making it 3 times cheaper than GPT-4!

• More Knowledgeable: with “April 2023” as a cutoff date!

• Longer Context Length: Up to 128k tokens! Making it the longest on the market, beating Anthropic’s Claude 2 and its 100k token context. 128k tokens is equivalent to about 100 pages!

• 

GPT-4-turbo is now available via a preview, limited version! You can already test it ; however, you’ll be limited to 100 calls per day for now. We’re expecting OpenAI to lift this limitation before end of 2023.

Side note: A faster and cheaper API means we've reached a viable GPT-4 option for businesses. I believe we'll have cheaper access by Q2 2024, making it the right time to experiment with GPT-4-Turbo, to be ready to scale by the next model/price update.

#2 GPTs: Build You Own GPT and monetize it through OpenAI’s GPTs Store!

• Empowering Customization and Creativity: OpenAI announced GPTs, a new ChatGPT feature that allows anyone, regardless of their coding expertise, to create custom versions of the popular ChatGPT, tailored for specific purposes​​​​. Whether you're a hobbyist looking to explore the realms of conversational AI or a business seeking to enhance your operations, GPTs open up a world of opportunities.

• Simplicity and Accessibility: The process is as straightforward as chatting with ChatGPT and detailing your requirements. For example, you could create a GPT trained on your personal recipe collection to assist in your culinary adventures, or one that helps in understanding complex data from your business files​​.

• The GPT Store: A Marketplace for Innovation: Perhaps the most exciting element is the introduction of the GPT Store. Reminiscent of Apple's App Store model, the GPT Store is a platform where creators can publish and monetize their GPTs. This marketplace not only offers a space for showcasing innovations but will also includes features like a search function and a leaderboard to highlight popular creations. OpenAI plans to reward creators of the most used and useful GPTs with a portion of their revenue, adding an incentive for quality and ingenuity​​.

• Future Possibilities and Monetization: The monetization strategy for GPTs is still evolving. Initially, OpenAI is focusing on a revenue-sharing model, with the possibility of introducing subscriptions for individual GPTs based on demand. The GPT Store is set to feature creations by verified builders, ensuring quality and reliability. This vetting process is crucial in maintaining a high standard for the GPTs available in the store​​.

Building your own GPT in a few clicks:

• NoCode Configuration: You can chat with ChatGPT to build your GPT or configure it yourself (“configure” tab)

• Knowledge and Actions: In the configure tab,

  • you’ll be able to add documents to be considered by your GPT.

  • you’ll also be able to allow your GPT to browse the web, generate images or use code interpreter

• Publishing: Once ready you can decide whether you want your GPT to be private, only accessible via a link or public!

Have Fun!

#3 Other Major Announcement

So much to cover! In a nutshell, assistant API, Dall.e.3, Copyright Shield, Speech-to-Text (Whisper) and Text-to-Speech (TTS) APIs!

What you need to know

1. Assistant API: This is a first step towards a streamlined Agent development. In a nutshell, it enable developing a full-fledged AI assistant based on your custom knowledge and connected to code interpreter!

   • Innovation and Competition: This new API competes directly with other AI companies, especially those helping you make your own AI assistants.

   • Custom Knowledge-Based Assistant: This API lets developers build AI assistants that are tailored to specific needs and can use unique, private information. This means you can create an AI assistant that knows exactly what your business needs, without dealing with embeddings and knowledge databases.

   • One-Stop Shop for AI Assistant: It's an all-in-one solution. Instead of using multiple tools, this API does it all - from following instructions to retrieving special information and understanding code. It's like a Swiss Army knife for creating AI assistants, making it easier and faster to develop them. We can expect this API to evolve into an Agent API in the coming year or so.

    

2. Dall.e 3 API: Midjourney’s competitor has now an API!
   OpenAI has introduced DALL·E 3 API, allowing developers to integrate advanced image generation capabilities directly into their apps. This API stands out for its built-in moderation tools and diverse format options, making it a powerful tool for creative and practical applications.
   Read more about the best image generation tools.

3. Whisper 3 (Speech-to-Text) API: The new version of Whisper features improved language recognition performance in a faster API!

4. TTS (Text-to-Speech) API: The newly introduced TTS API allows for the generation of human-like speech from text, offering multiple voice options and model variants. This development represents a leap forward in making AI-generated speech more natural and versatile.

5. Copyright Shield: Last but not least! A significant addition, OpenAI's Copyright Shield offers legal protection and cost coverage for developers against copyright infringement claims. This highlights OpenAI's commitment to safeguarding its users and their creations.

It feels like an early AI Christmas!

This is it for Today!

Until next time, this is Wassim Jouini, signing off. See you in the next edition!

Have a great Sunday and may AI always be on your side!