AIR#88 - May 31, 2024
Good morning, AI enthusiasts! As you sip your morning coffee, get ready to dive into today's edition of AIR: The AI Recon, where the world of artificial intelligence is buzzing with stories that are as fascinating as they are thought-provoking. Leading the headlines is a bombshell revelation from ex-OpenAI board member Helen Toner, who claims that Sam Altman's constant lies created a toxic culture, ultimately leading to his firing and subsequent rehiring. This scandal sheds light on the tumultuous dynamics within one of the most influential AI companies.
But it's not all about corporate drama. Dive into the exciting advancements in AI with the introduction of 1-bit language models. These smaller, faster models promise to slash AI energy use while maintaining nearly the same level of accuracy as their larger counterparts. This breakthrough could revolutionize the AI landscape by making powerful AI tools more accessible and environmentally friendly. Meanwhile, OpenAI's recent disruption of five covert influence operations using AI highlights the ongoing battle against deceptive uses of technology, emphasizing the importance of transparency and safety in the digital age.
As we explore these stories, let's not forget the broader implications they bring to the table. From Anthropic's innovative governance structure designed to avoid the pitfalls faced by OpenAI, to the ethical concerns raised by The Atlantic's partnership with OpenAI, today's edition is packed with insights that challenge our understanding of AI's role in society. Whether it's the promise of new technologies or the ethical dilemmas they pose, there's plenty to ponder as we delve into the dynamic world of artificial intelligence together. So, sit back, sip your coffee, and let's explore these captivating narratives.
Business
Sam Altman's Lies Created Toxic Culture, Says Ex-OpenAI Board Member
Ex-OpenAI board member Helen Toner claims Sam Altman's lies created a toxic culture, leading to his firing and later rehiring.
Disrupting Deceptive AI Uses by Covert Influence Operations
OpenAI disrupted 5 covert influence operations using AI, preventing significant audience engagement and enhancing transparency and safety.
A Devil's Bargain with OpenAI
The Atlantic partners with OpenAI, allowing ChatGPT to reference its content, sparking ethical concerns about AI's impact on journalism.
Anthropic's Governance Structure Aims to Avoid OpenAI's Pitfalls
Anthropic's unique governance aims to avoid OpenAI's pitfalls, prioritizing AI safety over profits while balancing investor interests.
US Delays Coal Plant Retirements Amid AI Power Surge
US delays coal plant retirements as AI power demand surges, straining grid and clashing with decarbonization targets.
Anduril Is Building the Pentagon's Deadly Drone Swarms
Anduril is developing advanced autonomous drone swarms for the Pentagon, revolutionizing aerial warfare with AI-driven systems.
OpenAI Reveals Russia and China Used Its A.I. for Covert Influence Campaigns
OpenAI reveals Russia, China, Iran, and Israel used its AI for covert influence campaigns, manipulating public opinion globally.
The ugly truth behind ChatGPT: AI is guzzling resources at planet-eating rates
AI like ChatGPT consumes massive resources, exacerbating environmental issues despite tech industry's net-zero claims.
The Atlantic Partners with OpenAI for Premium News Integration
The Atlantic partners with OpenAI to integrate its premium news into ChatGPT, enhancing discoverability and shaping future news experiences.
Apple's AI Servers to Use 'Confidential Computing' for Enhanced Privacy
Apple's AI servers will use 'confidential computing' to process user data while maintaining privacy, enhancing security measures.
Microsoft CEO Satya Nadella Worried About OpenAI-Apple Deal
Microsoft's CEO Satya Nadella is worried about an OpenAI-Apple deal affecting Microsoft's product ambitions.
Largest-Ever Botnet Operation Disrupts Global Malware Ecosystem
Europol's Operation Endgame disrupts largest-ever botnet, arrests key suspects, and seizes over 100 servers and 2,000 domains globally.
Engineering
π₯ βImpreciseβ Language Models: Smaller, Faster, and Nearly as Accurate
1-bit language models slash AI energy use, offering smaller, faster, and nearly as accurate alternatives to current large models.
[Paper] NPGA: Neural Parametric Gaussian Avatars β High-Fidelity Digital Faces
NPGA creates high-fidelity, controllable digital avatars from multi-view videos, outperforming previous methods in realism and animation accuracy.
All ChatGPT Free Users Get Browse, Vision, Data Analysis, File Uploads, and GPTs
All ChatGPT Free users now have access to browse, vision, data analysis, file uploads, and GPTs.
π₯ Better RAG with Reciprocal Rank Fusion & Hybrid Search
Assembled improves RAG accuracy using Hybrid Search with Reciprocal Rank Fusion, merging vector and keyword searches for better support query results.
[Paper] Optimal LLM Quantization with picoLLM by Picovoice
Picovoice's picoLLM optimizes LLM quantization, enhancing performance on various devices and reducing costs, surpassing existing methods.
Hacker Unleashes Jailbroken "Godmode" ChatGPT
Hacker releases jailbroken "Godmode" ChatGPT, bypassing OpenAI's guardrails. OpenAI swiftly takes action. The cat-and-mouse game continues.
PicoLLM: Cross-Platform On-Device LLM Inference Engine by Picovoice
Picovoice's picoLLM enables efficient on-device LLM inference across multiple platforms, optimizing performance on low-end hardware.
LLM-Powered Proxy Scraper Converts Websites to JSON via API
New LLM-powered API converts URLs to JSON effortlessly. Paste a link, get structured data. DM Blair Dulder for access. π
[GitHub] CinEx: FFmpeg Video Editing with LLMs in Elixir
CinEx: Simplify FFmpeg video editing with Elixir & LLMs in a single Livebook file. Easy setup with OpenAI adapter support. π
[Paper] FNet: Token Mixing with Fourier Transforms
FNet replaces self-attention in Transformers with Fourier Transforms, achieving 92-97% BERT accuracy while training up to 80% faster.
Academic
OpenAI Launches ChatGPT Edu for Universities
OpenAI launches ChatGPT Edu, an affordable AI tool for universities, enhancing learning, research, and campus operations with GPT-4o.
[Textbook] Deep Learning: Foundations and Concepts by Christopher M. Bishop and Hugh Bishop
"Deep Learning: Foundations and Concepts" by Chris and Hugh Bishop offers a comprehensive intro to key deep learning ideas for all learners.