weekend ai reads for 2023-12-01

📰 ABOVE THE FOLD: THE AFTERMATH

• The interested normie’s guide to OpenAI drama — Who is Sam Altman? What is OpenAI? And what does this have to do with Joseph Gordon-Levitt's wife??   Max Read, Substack

• The OpenAI saga isn’t over just yet — A new board and a promised investigation could threaten Altman’s happy ending   Platformerand

• Interview: Sam Altman on being fired and rehired by OpenAI The Verge

  • he also calls the Q* model breakthrough an “unfortunate leak” so a possible non-confirmation confirmation?

  • mostly evasive on details; good PR coaching

• A.I. Belongs to the Capitalists Now — The fight over OpenAI was at least partly about dueling visions of artificial intelligence. One side clearly won out.   New York Times

📻 QUOTE OF THE WEEK

“AI: The Coming Revolution,” Coatue, slide 34 (source)

🏗️ FOUNDATIONS & CULTURE

• Is My Toddler a Stochastic Parrot? — The world is racing to develop ever more sophisticated large language models while a small language model unfurls itself in my home.   The New Yorker

  • well-illustrated to boot

• Bill Gates teases the possibility of a 3-day work week where ‘machines can make all the food and stuff’ Fortune

• The Case for Using AI to Log Your Every Living Moment — Sam Liang, CEO of Otter, argues that life would be better if algorithms logged every spoken word so life events past can be lived and explored again.   Wired

  • no

  • related, AI Assistants Need to Know a Lot About You to Work Best. Is That OK? Cnet

    • also no

• Reshaping the tree: rebuilding organizations for AI — Technological change brings organizational change.   One Useful Thing (Ethan Mollick), Substack

  • Ethan Mollick seems to be moving to working in controlled (idealistic?) conditions; still worth a read but perhaps with a pinch or two of skepticism

• This company is building AI for African languages — AI models can’t understand African languages. Lelapa AI is trying to change that.   MIT Technology Review

• I’ve spent the past 19 months knee-deep in generative AI, and I think I’m ready to offer three predictions about what AI will mean for typeface design. Jonathan Hoefler, Threadsand

  • related perspective, If designers don't embrace AI the world “will be designed without them” says AirBnb founder Dezeen

🎓 EDUCATION

• Should Universities Build Their Own Custom AI Tools in House? Government Technology

  • easy for well-heeled institutions; what about the other 99% ?

• Exclusive: OpenAI wants ChatGPT in classrooms Reuters

• $400k graduate salaries and gun-for-hire ‘SWAT teams’ — inside the wild AI talent market of 2023 Sifted

• Carnegie Learning Announces LiveHint AI — The First Generative AI Math Tutor Trained to Think Like a Student   press release

• This School Leader Believes AI Could Transform Education for Students With Dyslexia Education Week

📊 DATA & TECHNOLOGY

• Beyond Guesswork: The Rise of Retrieval Augmented Generation Art of Saience

• Removing RLHF Protections in GPT-4 via Fine-Tuning arXiv

  • public APIs are often easy to exploit and LLM APIs are no exception, apparently

• How Much Does it Cost to Use an LLM? Thomasz Tunguz

  • analysis done prior to the price reduction Open AI announced but still interesting analysis

  • would like to see a total cost of ownership analysis, particularly for open source models

• Watermarks in the Sand: Impossibility of Strong Watermarking for Generative Models arXiv

• LLMs cannot find reasoning errors, but can correct them! arXiv

• Request for proposals: benchmarking LLM agents on consequential real-world tasks Open Philanthropy

🎉 FUN and/or PRACTICAL THINGS

• AI Garage Sale — haggle with AI to buy real products

  • they did a great job mimicking the experience so it is as annoying as real-life

• Several popular AI products flagged as unsafe for kids by Common Sense Media Tech Crunch

  • possibly related, Meta disbanded its Responsible AI team — A new report says Meta’s Responsible AI team is now working on other AI teams.   The Verge

• Unauthorized “David Attenborough” AI clone narrates developer’s life, goes viral Ars Technica

• Sports Illustrated Published Articles by Fake, AI-Generated Writers — We asked them about it — and they deleted everything.   Futurism

• DepolarizingGPT — The 3-Answer Political AI from Developmental Politics

  • ask a question and get left-wing, depolarizing, and right-wing responses

  • limited number of tokens in “Quick Trial” but enough to get a sense of how it works

• Users Can't Speak to Viral AI Girlfriend CarynAI Because CEO Is in Jail for Arson 404 Media ($)

  • what a wild headline

• Track AI Answers — Monitor and track online reputation of people, brands and products in AI answers

  • the most negative fact Google Bard AI knows about George Clooney is “He has been nominated for seven Primetime Emmy Awards, and has won three.”

• A Spanish agency became so sick of models and influencers that they created their own with AI—and she’s raking in up to $11,000 a month Fortune

• Stable Diffusion XL Turbo Beta — Real-Time Text-to-Image Generation

  • requires registration

  • real-time image generation

  • doesn’t do great with faces or hands; pretty good with food

• AI Splash — The best free stock photos and royalty-free images... powered by robots everywhere.

  • some are rather nice; which is a commentary on prompt engineering and image curation, not the technology’s ability to generate “nice” images

• Christopher Nolan on the Promise and Peril of Technology The Atlantic ($)

🧿 AI-ADJACENT

• The Information Is Beautiful Awards 2023 Data Visualization Society

  • always fun; usually important