- That AI Thing
- Posts
- weekend ai reads for 2023-08-31
weekend ai reads for 2023-08-31
📰 ABOVE THE FOLD: GOOGLE CLOUD NEXT
Musk, tech CEOs to attend Schumer’s AI Senate forum The Hill
Google DeepMind has launched a watermarking tool for AI-generated images — It’s the first Big Tech firm to publicly launch one, after a group of them pledged to develop them at the White House in July. MIT Technology Review
Google Meet’s new AI will be able to go to meetings for you The Verge
Well, it summarizes the meeting and notes action items
Google Gemini Eats The World – Gemini Smashes GPT-4 By 5X, The GPU-Poors Semianalysis
Google to Add AI Models from Meta, Anthropic to Its Cloud Platform Bloomberg
📻 QUOTE OF THE WEEK
The way I see it, Character.AI brings our beloved characters closer to us. This is a process that I often describe as affective reception, fleshing out the emotions that we feel for characters and fiction.
Nicolle Lamerichs (source)
🏗️ FOUNDATIONS & CULTURE
Web Scraping for Me, But Not for Thee Technology & Marketing Law Blog
related, Generative AI and intellectual property Benedict Evans
Hallucinations for Fun and Profit (20:14) Tim Hwang, YouTube
The Next Wave of Scams Will Be Deepfake Video Calls From Your Boss Bloomberg
Behind the AI boom, an army of overseas workers in ‘digital sweatshops’ Washington Post
Focuses on Scale AI (valuation: $7B) reportedly paying its Filipino “taskers” less than minimum wage, $6-10 per day.
Watch this video featuring his top 5 tips for prompting, then try them yourself at http://claude.ai. (3:35) Anthropic, Twitter (sorry)
Autonomous AI Founder: Can AI Be A Startup Founder AI Super Founder, Substack
UAE launches Arabic large language model in Gulf push into generative AI Financial Times
🎓 EDUCATION
University of Michigan to provide custom AI tools to campus community The Detroit News
Prompt and Process with Ethan Mollick (55:37) Possible Podcast
AI in Education Attracts Startups and Cash (12:48) Tech News Briefing, Wall Street Journal
Perception, performance, and detectability of conversational artificial intelligence across 32 university courses Scientific Reports, Nature
📊 DATA & TECHNOLOGY
Why host your own LLM? Andrew Marble
Llama 2 is about as factually accurate as GPT-4 for summaries and is 30X cheaper Anyscale
Supercharging ML/AI Foundations at Instacart Tech at Instacart
Details of ML investments Instacart has made to make the user experience more personalized and establish a foundation to incorporate LLMs into their systems
Here’s what the authors of LLaMA have to say about why RLHF is so important… Cameron R. Wolfe, Twitter (sorry)
Faith and Fate: Limits of Transformers on Compositionality arXiv
Analysis of the errors made by LLMs on compositional tasks gives evidence that these LLMs might be mostly memorizing training data, and not really learning multi-step algorithms to solving the problems.
🎉 FUN and/or PRACTICAL THINGS
Prompt Pal Be more productive at work and in your day-to-day life with the industry's best AI prompts
AI Dad Jokes Hugging Face
Upload an image and get some “dad jokes” which seems to mean unfunny observations about the image
Film Finder — Film Recommendation using GenAI
Exactly What Is Copy, the First AI-Powered Fashion Magazine, Trying to Prove? Vogue
spoiler: The author prefers a real financial advisor; she also happens to be a former financial advisor; one of her valid gripes is with ChatGPT’s limits with math
Primarily focused on predicting rain, “powered by AI technology.” Jury is out whether it’s a replacement for the great Dark Sky (RIP)
Call of Duty partners with Modulate to use AI to fight toxicity in voice chat Venture Beat
Midlibrary The most advanced library of genres, artistic movements, techniques, titles, and artists' styles for Midjourney AI
Lots of examples of the evolution of the 3,600+ (!) styles available in Midjourney
🧿 AI-ADJACENT
The Usefulness of a Memory Guides Where the Brain Saves It Quanta Magazine