Good morning. It’s Monday, July 1st.

Did you know: 45 years ago today, the first Sony Walkman was released?

In today’s email:

• Apple + Gemini Deal

• New Smart Glasses

• Vision Model Ranking

• Robotics Development

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Apple could announce a Google Gemini deal this fall

Apple is poised to expand its AI integration beyond iPhones, iPads, and Macs to include Vision Pro headsets, according to Bloomberg's resident Apple whisperer Mark Gurman. This move will likely incorporate Google Gemini alongside ChatGPT, enhancing Apple devices with advanced chatbot capabilities. While Apple Intelligence is slated for beta release this fall, it aims to monetize AI directly, moving beyond hardware-centric features. Gurman suggests a potential shift to subscription-based Apple Intelligence services in the future.

Despite delays in Vision Pro integration this year, Apple continues to refine its retail strategy, allowing users to preview personal media and introducing a more comfortable Dual Loop headband. Analyst Ming-Chi Kuo predicts AirPods with infrared cameras by 2026, enhancing spatial audio and gesture controls, particularly when paired with the Vision Pro. Read more.

The Ray-Ban Meta Smart Glasses Are About To Get New Competition

Solos, known for its AirGo smart glasses, is gearing up to compete directly with Ray-Ban Meta by introducing the AirGo Vision model. This new iteration integrates a front-facing camera positioned discreetly in the frame's corner, similar to Ray-Ban's popular models, and promises easy access to AI-powered visual search and interactive features on the move.

Unlike its competitors, Solos embraces an open architecture, accommodating AI models like ChatGPT-4o, Google Gemini, and Anthropic Claude. The camera, situated on the glasses' arm rather than within the main frame, supports voice-activated photo capture but excludes video recording.

AirGo Vision's modular design allows users to swap out the front panel for personalized aesthetics. Adding to its functionality, the glasses feature an LED notification light for alerts and camera status updates. Solos plans to launch the AirGo Vision later this year alongside three camera-free styles priced at $250. Read more.

GPT-4o and Claude 3.5 Sonnet dominate vision language models  

In a recent analysis conducted by LMSYS Org, GPT-4o and Claude 3.5 Sonnet emerged as leaders in vision language models (VLMs), excelling notably in image recognition compared to rivals like Gemini 1.5 Pro and GPT-4 Turbo. Over a span of two weeks, feedback from more than 17,000 users in 60 languages underscored these models' superior performance. While Claude 3 Opus showed strong capabilities in linguistic tasks, Gemini 1.5 Flash demonstrated comparable effectiveness in VLM applications.

The study encompassed a wide array of practical uses such as image description, mathematical problem-solving, document comprehension, meme interpretation, and storytelling. Looking ahead, LMSYS Org plans to enhance its platform to accommodate multiple images, PDFs, videos, and audio files. Read more.

MIT’s Soft Robotic System is Designed to Pack Groceries

MIT's CSAIL department has introduced RoboGrocery, a soft robotic system designed for packing groceries. Combining computer vision and a flexible gripper, it efficiently handles various items. In tests, researchers presented 10 unfamiliar objects, ranging from delicate grapes to sturdy soup cans, on a conveyor belt.

The system's vision system identifies objects and assesses their placement. For instance, delicate items like grapes are handled carefully, while rigid items like soup cans are placed securely in bags. Lead author Annan Zhang notes the system's potential for automating grocery packing, although commercial deployment is pending further development. Beyond groceries, the technology shows promise for industrial applications like recycling plants. Read more.

Etcetera: Stories you may have missed

Microsoft’s AI boss thinks it’s perfectly OK to steal content if it’s on the open web

Researchers develop the fastest possible flow algorithm

MIT robotics pioneer Rodney Brooks thinks people are vastly overestimating generative AI

ChatGPT is hallucinating fake links to its news partners’ biggest investigations

AI deals between Microsoft and OpenAI, Google and Samsung, in EU crosshairs

Apple Silicon exec joins Rain AI to develop new hardware

Amazon beefs up AI development, hiring execs from startup Adept and licensing its technology

Baidu launches upgraded AI model, says Ernie Bot hits 300 million users

Inline ChatGPT integrates across all apps, replacing selected text with AI-generated outputs via ⌘-Shift-1. Free with OpenAI API key.

Plus AI PowerPoint Maker creates professional PowerPoint presentations in minutes using AI, integrating directly with PowerPoint for compatibility and ease of use.

ConsoleX.ai is a unified LLM playground featuring AI chat interfaces, LLM API playground, and batch evaluation, supporting all mainstream LLMs.

La Growth Machine automates personalized, multi-channel conversations across LinkedIn, Email, Voice, Calls, and X. Includes CRM integration.

Hedra is a platform for developing foundation models, enabling the creation of virtual worlds, characters, and narratives with complete creative control.

📄 Can LLMs Learn by Teaching? A Preliminary Study

📄 Scaling Synthetic Data Creation with 1,000,000,000 Personas ​​

📄 HuatuoGPT-Vision, Towards Injecting Medical Visual Knowledge into Multimodal LLMs at Scale  

📄 AutoRAG-HP: Automatic Online Hyper-Parameter Tuning for Retrieval-Augmented Generation

📄 Direct Preference Knowledge Distillation for Large Language Models  

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, apply here.