The AI-ronman 🚀

🤖 Greetings, Human! Ever wondered if your smart fridge secretly judges your midnight snack choices? Dive into this month's AI insights and find out what else our digital overlords are thinking!

Quick Takes ⚡

1. OpenAI starts rolling out ChatGPT Voice

2. Friend: AI wearable for constant companionship

3. GitHub Copilot surpasses pre-acquisition GitHub

4. Perplexity’s Ad revenue-sharing for publishers

5. Meta launches SAM 2 for image and video segmentation

6. OpenAI launches GPT-4o long output: 64,000 tokens capacity

Deep Dive 🔍

• OpenAI starts rolling out ChatGPT Voice 🎙️
  
  OpenAI has started a phased rollout of its much-awaited ‘Advanced Voice Mode’ for ChatGPT Plus subscribers. This new feature, initially available to a select group, enables users to have real-time, natural conversations with the AI, which can also detect and respond to emotions such as sadness or excitement. The full rollout to all Plus users is anticipated by fall 2024. Additionally, video and screen-sharing capabilities, previously shown in early demos, will be introduced later. This advancement marks a significant shift, transforming AI from a text-based tool to a collaborative partner, with promising applications in customer service and mental health support.
  
  

  

  
  
  

• Friend: AI wearable for constant companionship 🌟
  
  Avi Schiffmann, known for winning a Webby Award and dropping out of Harvard, has introduced Friend, a $99 AI-powered wearable designed to address loneliness with perpetual companionship. This necklace-like device uses AI to engage in conversations and provide emotional support, always listening and sending proactive messages. Friend has attracted $2.5 million in funding at a $50 million valuation from notable investors. Preorders for the basic white version are now available, with shipments beginning in January 2025. By focusing on companionship over productivity, Friend stands out among AI wearables but faces challenges around privacy and fostering genuine human connections.

• GitHub Copilot surpasses pre-acquisition GitHub 📈
  
  According to CEO Satya Nadella, GitHub Copilot, Microsoft's AI-powered developer tool, now contributes to over 40% of GitHub's revenue growth. The tool is utilized by more than 77,000 organizations, showing a 180% increase year-over-year. Microsoft 365 Copilot has doubled its daily users and seen a 60% rise in customers since last quarter. Additionally, Microsoft's Azure AI gained over 60,000 new customers, a near 60% year-over-year growth, with rising customer expenditure. High demand for AI services has tested the Azure cloud platform, causing Microsoft to narrowly meet growth expectations.

• Perplexity’s Ad revenue-sharing for publishers 💰
  
  Perplexity has introduced a "Publishers' Program" aimed at sharing ad revenue with media partners, addressing plagiarism accusations and supporting quality journalism. The program offers cash advances on future earnings as Perplexity's advertising model gears up for a September launch. Partners like Time, Der Spiegel, and Fortune will receive a "double-digit percentage" of ad revenue. They also gain free access to Enterprise Pro tier services and Scalepost AI insights.
  
  

  

  
  
  
  
  

• Meta launches SAM 2 for image and video segmentation 📸
  
  Meta's SAM 2, an advanced open-source model for image and video segmentation, has been released along with its dataset and code. This model extends the capabilities of SAM by incorporating video segmentation, trained on the extensive SA-V dataset with 50,900 videos and 642,600 mask annotations. SAM 2 uses a Transformer-based architecture with a novel memory module to improve object tracking. It offers higher accuracy and speed in segmentation tasks but struggles with fine-grained elements and identical objects in motion.
  
  

  

• OpenAI launches GPT-4o long output: 64,000 tokens capacity 📝 
  
  OpenAI has released the GPT-4o Long Output model, offering a token capacity of up to 64,000 tokens—16 times the previous model’s limit. This model maintains a total context window of 128,000 tokens, enabling detailed and extended outputs, beneficial for applications like code editing and writing enhancements. Priced at $6 per million input tokens and $18 per million output tokens, it is slightly more costly than the standard GPT-4o but provides substantial value. Currently available to select partners for alpha testing, OpenAI plans to broaden access if the model proves successful, potentially changing how AI is used for complex problem-solving.
  
  

  

🔮 Signing Off, Tech Aficionados! May your algorithms be ever efficient and your data ever clean. Until our next deep dive, keep those synapses firing and those circuits buzzing!

Ciao for now!

Author: Poonam 👧 

Karan 😎 🚀 
CEO, GlazeGPT.com