🗞 OpenAI Unveils Voice-Enabled ChatGPT with Advanced Features 🗣️

🔎 The Latest on the AI Frontier:

• OpenAI Unveils Voice-Enabled ChatGPT with Advanced Features 🗣️

• Microsoft's Azure Cloud Growth Slows, Disappoints Investors 📉

• Meta AI Provides Inconsistent Info on Trump Assassination Attempt 🤖

• Nvidia CEO Envisions Three-Computer System for Future AI 🖥️

• Wix Launches AI-Powered Blog Creation Tool for SEO ✍️

• AI Pendant 'Friend' Aims to Combat Loneliness for $99 📿

• AI Writing in Scientific Papers Raises Plagiarism Concerns 🔍

• MIT's 'Thermometer' Method Enhances AI Model Calibration 🌡️

• Study Shows AI Cannot Replace Human Teachers in Education 🎓

🗣️ OpenAI launches voice-enabled ChatGPT with advanced conversational capabilities. Link

• Initially available to a "small group of users" paying for ChatGPT Plus, expanding to all Plus subscribers ($20/month) this fall.

• Powered by new GPT-4o model combining voice, text, and vision capabilities.

• Tested with over 100 external red-teamers speaking 45 languages from 29 countries for safety and quality.

📉 Microsoft's AI-powered cloud business Azure disappoints in latest earnings report. Link

• Azure grew 29% last quarter, below analyst expectations and down from 31% growth in the previous quarter.

• Microsoft's total cloud revenue was $28.5 billion, up 19% but still disappointing investors.

• Company spent $19 billion on capex, up 78% year-over-year, largely on AI initiatives.

🤖 Meta AI tool provides inconsistent information about Trump assassination attempt. Link

• Meta's AI assistant called the Trump assassination attempt "fictional" when directly asked about it.

• The tool provided detailed information about Kamala Harris' 2024 campaign but little on the Trump incident.

• Meta acknowledged its AI is providing "incomplete, inconsistent, or out of date information" and is working on a fix.

🖥️ Nvidia CEO predicts AI will require "three computers" in the future. Link

• Huang says we'll need 3 distinct computers for AI: one to create, one to simulate, and one to run it.

• Nvidia has raised $2.5 billion in funding at a $50 million valuation for AI initiatives.

• The CEO spoke for an hour-long interview at the Siggraph conference, focusing heavily on AI's future.

✍️ Wix introduces AI-powered blog post creation tool for SEO optimization. Link

• New feature can generate entire blog posts including imagery, aiming to boost organic traffic by 86% on average.

• Tool suggests topics, titles, and SEO keywords while allowing human writers to edit and maintain content authenticity.

• Raises concerns about potential proliferation of AI-generated content at the expense of genuinely useful information online.

📿 AI-powered pendant 'friend' aims to combat loneliness for $99. Link

• Wearable device responds to voice input with text messages, offering companionship without traditional AI assistant functions.

• Set to begin shipping in 2025, the always-listening pendant uses end-to-end encryption and allows one-click memory deletion for privacy concerns.

• Raised $2.5 million in funding at a $50 million valuation, joining the growing market of AI-powered wearable communication devices.

🔍 AI-assisted writing in scientific papers raises concerns about plagiarism and ethics. Link

• Study estimates 10% of biomedical abstracts in early 2024 used AI language models, equivalent to 150,000 papers per year.

• Journals are updating policies to allow some AI use with full disclosure, but clearer guidelines are urgently needed.

• AI detection tools struggle with accuracy, especially for lightly edited text, raising risks of false accusations.

🌡 MIT's "Thermometer" method improves AI model calibration efficiency. Link

• Researchers claim well-calibrated models with blogs get 86% more organic traffic compared to sites without blogs.

• Technique works on LLMs with billions of parameters, efficiently calibrating them for new tasks.

• Aims to help users know when to trust a model's predictions, potentially preventing deployment failures.

🎓 MIT experiment shows AI cannot replace human teachers in education. Link

• One study analyzed 424 experimental studies on social and emotional learning (SEL) programs.

• Los Angeles Unified School District, the 2nd largest in the U.S., recently banned smartphone use in classrooms.

• In 1984, psychologist Benjamin Bloom posed the "2-sigma problem", highlighting that individual tutoring is the best educational paradigm.

Have any feedback? Send us an email