AIR#103 - June 15, 2024
Good morning, AI aficionados! Grab your coffee and get ready to dive into today's edition of AIR: The AI Recon. Leading the headlines is Microsoft's decision to delay the release of its highly anticipated Recall AI feature due to privacy concerns. Instead, the tech giant is opting for a limited preview through the Windows Insider Program. This cautious approach underscores the growing importance of security in AI advancements, making it a must-read for anyone keeping an eye on corporate responsibility in tech.
Meanwhile, NVIDIA is making waves with the launch of Nemotron-4 340B, an open model designed for generating synthetic data to enhance LLM training across industries. Available on Hugging Face, this innovation promises to revolutionize data generation, offering developers a powerful tool to push the boundaries of what's possible. And if you're curious about the costs of self-hosting, a new report reveals that running Llama-3 8B-Instruct can be significantly cheaper than using ChatGPT, especially with the right hardware. It's a fascinating dive into the economics of AI that business-minded readers will find particularly enlightening.
But it's not all about the big players. SoftBank's latest AI development is turning angry customer calls into calm, soothing interactions, reducing stress for call center operators and improving customer service experiences. This practical application of AI technology highlights the everyday benefits that AI can bring, making it a heartwarming read amidst the high-tech innovations. Whether you're here for the corporate drama, groundbreaking tech, or practical applications, today's edition is packed with stories that will both intrigue and challenge you. So, sit back, sip your coffee, and let's delve into the dynamic world of artificial intelligence together!
Business
🔥 Microsoft Delays Recall AI Release Over Security Concerns
Microsoft delays Recall AI release due to privacy concerns, opting for limited preview with Windows Insider Program instead.
🔥 Cost of Self-Hosting Llama-3 8B-Instruct
Self-hosting Llama-3 8B costs $17 per 1M tokens vs. ChatGPT's $1. Self-hosting hardware can cut costs to <$0.01 per 1M tokens over 5.5 years.
SoftBank's AI Makes Angry Customers Sound Calm on Phone
SoftBank's AI transforms angry customer calls into calm tones, reducing stress for call center operators and enhancing customer interactions.
OpenAI Adds Former NSA Chief to Board
OpenAI adds ex-NSA chief Paul Nakasone to its board to enhance AI-driven cybersecurity and joins Apple for ChatGPT-Siri integration.
Clearview AI Faces Class-Action Settlement: You Could Get a Stake
Clearview AI settles privacy lawsuit, offering a 23% stake in the company to those whose faces were used in its database, pending court approval.
Engineering
🔥 NVIDIA Unveils Nemotron-4 340B for Synthetic Data Generation
NVIDIA launches Nemotron-4 340B, an open model for generating synthetic data, enhancing LLM training across industries. Available on Hugging Face.
🔥 Apple's AI Strategy: Core Model Performance and Beyond
Apple's AI strategy focuses on personalized, on-device models, enhancing user experience while maintaining privacy.
🔥 [GitHub] NVIDIA Warp: High-Performance Python Framework for GPU Simulation and Graphics
NVIDIA releases Warp, a Python framework for high-performance GPU simulation and graphics. Ideal for physics, robotics, and ML pipelines!
[Video] Developing an LLM: Building, Training, Finetuning (Sebastian Raschka)
Sebastian Raschka's new video explains building, training, and finetuning large language models (LLMs) from scratch. Check it out!
Academic
Turning the Tables on AI: Using AI to Think More, Not Less
Use AI to enhance thinking, not replace it. Let AI ask questions, edit, and critique to boost creativity and ownership in your writing.