AIR#62 - May 06, 2024
Good morning, AI enthusiasts! As you enjoy the first rays of sunlight with your coffee in hand, today's edition of AIR: The AI Recon is here to sprinkle your day with the most enthralling AI stories that are buzzing across the globe. From Microsoft CTO's revelations on OpenAI that have set the internet abuzz to Meta's eye-watering $30 billion GPU splurge that's making the Manhattan Project look like pocket change, we're diving deep into the heart of AI's financial and technological whirlwind. And for those who marvel at the blend of technology and creativity, Randy Travis's comeback with an AI-cloned vocal song is a testament to the boundless possibilities AI holds for the music industry.
But that's not all. The tech world is also abuzz with innovations that are pushing the boundaries of what we thought was possible. Skeeto's hash-prospector is revolutionizing how we discover and optimize integer hash functions, making waves among developers for its groundbreaking approach. Meanwhile, the realm of AI research is not standing still, with Stanford's exploration of machine unlearning and the Infini-Gram's leap in scaling n-gram models to a trillion tokens, reshaping our understanding of text analysis and AI's capabilities.
As we navigate through these stories, let's not forget the broader implications and the ethical considerations they bring to the forefront. From the awe-inspiring to the controversial, each piece offers a glimpse into the multifaceted world of AI, challenging us to think about the future we're building. So, as you take another sip of your coffee, let these tales of innovation, challenge, and breakthrough inspire you to ponder the role of AI in shaping our world. Here's to a day filled with curiosity and discovery in the ever-evolving landscape of artificial intelligence. Let's dive in!
Business
🔥 Microsoft CTO Shares Insights on OpenAI in Internal Tech Emails
Microsoft CTO shares insights on OpenAI, sparking widespread interest and discussion online.
🔥 Meta's GPU Spending Rivals Manhattan Project Costs, Says Ethan Mollick
Meta's GPU investment nears Manhattan Project levels, spending $30 billion on NVIDIA GPUs for AI, outpacing Apollo mission costs.
Randy Travis Releases AI-Cloned Vocal Song After Stroke
Randy Travis releases "Where That Came From," his first AI-cloned vocal song since his 2013 stroke, marking a new era in music technology.
ConversAItions: The AI Copilot Transforming Meetings with Real-time Prompts, Fact-checking, and Note-taking
ConversAItions, the AI copilot, revolutionizes meetings with real-time prompts, fact-checking, and note-taking for all teams.
Exploring the Vast Potential of Niche AI Applications
Niche AI applications are booming, proving foundational models like ChatGPT won't make them obsolete. The future is bright for Applied AI, requiring deep integration and specialized solutions.
"Introducing Milio: The AI Interview Coach That Provides Real-Time Answers"
Ace your job interviews with Milio, the real-time AI coach for everything from Leetcode puzzles to behavioral questions. Just talk, and Milio guides you instantly. 🎧🚀 $35/month for unlimited access.
Sam Altman's Vision of AI Regulation in Early Email to Elon Musk
Altman's early vision for AI with Musk: prioritize safety, foundation ownership, and global good, now contrasts with OpenAI's Microsoft deal.
HorseGPT's Derby Predictions Fall Short: AI Can't Outrace the Odds
HorseGPT, aimed to predict Kentucky Derby winners, flopped, showing AI can't always beat the unpredictability of horse racing.
Travel Companies Use ChatGPT to Craft Apologies for Ruined Vacations
Travel companies now use ChatGPT to craft apologies for ruined vacations, turning tedious tasks into efficient, emotion-free responses.
Engineering
🔥 [Github] skeeto/hash-prospector: Discover Integer Hash Functions Automatically
Discovering next-gen integer hash functions with skeeto/hash-prospector: a tool that auto-generates, tests, and optimizes with AI, pushing boundaries in hashing tech.
🔥 [Github] Deep Reinforcement Learning: Zero to Hero by alessiodm
From novice to expert, alessiodm's GitHub guides you through deep reinforcement learning with hands-on Atari and Moon landing AIs.
🔥 [Paper] SEQUOIA: Enabling Real-Time Llama2-70B Inference on RTX4090 with Sub-Second Latency
Sequoia framework slashes LLM inference time on GPUs like RTX4090 to sub-second, making real-time AI content creation a reality.
[Github] LFG: CLI Tool for Finding Terminal Commands with llama3
Discover LFG: The CLI tool revolutionizing terminal command searches with llama3, making tech tasks easier. 🦙
[Github] AviSoori1x's seemore: DIY Grok 1.5/GPT-4 Vision Model in PyTorch
AviSoori1x launches seemore: A DIY Grok 1.5/GPT-4 Vision Model in PyTorch, blending cutting-edge AI for visual and language understanding.
AI Revolutionizes Game Development: Lower Costs, Enhanced Creativity
AI reshapes game development: slashing costs, boosting creativity, and personalizing player experiences, while humans remain essential for true innovation.
TensorDock's Ultimate Guide to Choosing the Best GPU for Machine Learning
Discover the best GPUs for machine learning: RTX & Ada for small models, RTX 6000 & L40 for medium, H100 for large. Cost-effective & powerful options detailed.
ChatGPT Creates Custom Mission Patches with Mission Patch GPT by amentum.io
ChatGPT teams up with amentum.io to craft personalized mission patches. Sign up required for this creative journey!
[Github] AI Robotics: Learn Football Tricks with Reinforcement Learning
Dive into AI robotics: Master football tricks with reinforcement learning - no GPU needed, perfect for hobbyists!
OpenAI's Hide and Seek Game Breakthrough [Video]
OpenAI's AI outsmarts the classic Hide and Seek game, showcasing surprising strategies and breaking norms! Watch the breakthrough 🤖🎮
Chris McCormick's LuaVST: Generating VST Plugins with ChatGPT
Chris McCormick's LuaVST lets you generate VST plugins with ChatGPT - a mix of productivity boost and creative aid, but complex tasks need human touch.
Academic
🔥 [Paper] Machine Unlearning in 2024 by Ken Ziyu Liu - Stanford Computer Science
Stanford's Ken Liu explores machine unlearning, focusing on removing undesired data from AI models without full retraining, touching on privacy, safety, and copyright issues.
[Paper] Infini-Gram: Unbounded n-Gram Models Scale to Trillion Tokens
Infini-Gram scales n-gram models to a trillion tokens, modernizing text analysis and enhancing neural LLMs with near-instant calculations.
Demystifying LLMs: A Non-Mathematical Explanation by Miguel Grinberg
Miguel Grinberg demystifies LLMs, explaining their operation without math, focusing on token predictions for text generation.
Genius Group Sparks Outrage with Alan Turing AI Chief Stunt
Genius Group faces backlash for naming an AI chatbot "Alan Turing" as their AI chief, sparking ethical debates and calls for an apology.
[NeurIPS 2023 Paper] PlanBench: Benchmarking LLMs on Planning and Reasoning Capabilities
PlanBench, NeurIPS 2023's new benchmark, tests LLMs on planning and reasoning, revealing gaps even in top models.