AI Weekly: OpenAI’s New Model, Apple’s AI Updates, and More!

This past week in AI saw two major announcements: OpenAI's release of the 01-Preview model and Apple's latest AI features unveiled at their iPhone event. Here’s a rundown of the key highlights from this action-packed week.

?? OpenAI’s 01-Preview Model

OpenAI recently launched the 01-Preview model, previously teased under the names Q* and Strawberry. OpenAI has reset its naming convention, moving away from the GPT series. OpenAI 1 marks the beginning of a new generation of models.

What makes 01 stand out? It’s designed for advanced reasoning, logic, and complex problem-solving. While GPT-4 is still recommended for most tasks, 01 excels in areas like advanced math and puzzles, thanks to its Chain of Thought prompting. This approach allows the model to think through problems step-by-step, resulting in slower but more precise responses.

There’s also the 01 Mini model, which offers a more affordable and faster alternative, optimized for STEM reasoning and priced 80% lower than the full 01-Preview model.

?? Apple’s AI-Powered iPhone 16 Event

Apple's recent event introduced some AI updates, many of which were previewed at WWDC but are now set to roll out with the iPhone 16. Notable AI features include:

• AI-driven photo editing that can remove objects from the background.
• Enhanced Siri interactions, where users can nod or shake their head to respond.
• Built-in AI-powered translation for the Apple Watch.
• Private Cloud Compute to run larger AI models securely.
• A new Visual Intelligence feature (expected in 2025) that will provide real-time information based on images taken with the iPhone.

Many of these features will start rolling out with iOS 18.1, while the more advanced AI tools will arrive over the coming months.

?? Adobe’s Firefly Text-to-Video

Adobe has launched its Firefly text-to-video generation tool, which aims to compete with Runway’s Gen-2. What sets Firefly apart is its focus on ethical sourcing, as the model is trained solely on openly licensed content and Adobe Stock. Previews of the generated videos are impressive, offering high-quality, short clips that demonstrate significant potential for commercial use.

?? Mistral’s Pixtral 12B Model: Open Source Multimodal AI

Mistral released the Pixtral 12B model, an open-source multimodal AI capable of accepting text and images as input. This model represents a major advancement, giving developers an opportunity to build customized AI-driven applications.

?? Roblox’s 3D AI Generation

Roblox has introduced plans for AI-generated 3D world creation. With this new feature, creators will be able to describe scenes in text, and AI will generate immersive environments, such as steampunk castles or Scottish Highlands landscapes. The goal isn’t to replace creativity but to empower more people to build dynamic, interactive worlds on the platform.

Final Thoughts

Between OpenAI’s groundbreaking 01-Preview model, Apple’s AI-enhanced hardware, and Adobe’s advancements in text-to-video generation, this week was packed with significant developments. AI tools are rapidly becoming more integrated into everyday tasks, from solving complex problems to creating immersive digital content.

Stay tuned for more updates next week. What excites you most about these AI innovations? Let us know in the comments!

Follow AI Weekly for more news, insights, and tools from the world of artificial intelligence.