- Chaos Theory
- Posts
- 🥟 Chao-Down #135 OpenAI shutters AI text detector over low accuracy, The labor shortage facing the US Semiconductor Industry, Wayfair shows off AI to reimagine interior design
🥟 Chao-Down #135 OpenAI shutters AI text detector over low accuracy, The labor shortage facing the US Semiconductor Industry, Wayfair shows off AI to reimagine interior design
Plus, a look at the underclass forming in AI - the army of contractors building ChatGPT, search engines, and other AI tools who say they are underpaid and mistreated.
I’ve talked in the past about the difficulty inherent in detecting AI-generated text and it looks like OpenAI is throwing in the towel on their efforts to do the same. Based on initial reporting by Decrypt, OpenAI quietly shut down its AI detection tool due to a low rate of accuracy.
Not only do these tools not work reliably, but they have been shown to discriminate against non-native English speakers, unfairly flagging their writing as being AI-generated.
As tools like ChatGPT become more ubiquitous, we’re bound to see a flood of more machine-created text floating around the internet.
The question is, are we prepared for that future?
-Alex, your resident Chaos Coordinator.
What happened in AI? 📰
OpenAI scuttles AI-written text detector over ‘low rate of accuracy’ (TechCrunch)
Chipping Away: Assessing and Addressing the Labor Market Gap Facing the U.S. Semiconductor Industry (Semiconductor Industry Association)
Cohere releases Coral, AI assistant designed for enterprise business use (VentureBeat)
America already has an AI underclass - Search engines, ChatGPT, and other AI tools wouldn’t function without an army of contractors. Now those workers say they’re underpaid and mistreated (The Atlantic)
Amazon wants Alexa to bring AI into the home (Axios)
Wayfair unveils generative AI solution to reimagine interior design (VentureBeat)
Always be Learnin’ 📕 📖
A First Course in Causal Inference (arxiv)
The Architect’s Guide to Storage for AI (The New Stack)
A developer's guide to prompt engineering and LLMs (The GitHub Blog)
How to Make a Great Conference Talk (switowski.com)
Do large language models know what they are talking about? - Stack Overflow Blog
What We Know About LLMs (Primer) (willthompson.name)
Projects to Keep an Eye On 🛠
alantech/marsha: Marsha is a functional, higher-level, English-based programming language that gets compiled into tested Python software by an LLM (Github)
Custom Open-Source LLMs (Scale AI)
shikras/d-cube: A detection dataset with class names characterized by intricate and flexible expressions. Paper: "Exposing the Troublemakers in Described Object Detection". (Github)
The Latest in AI Research 💡
RecallM: An Architecture for Temporal Context Understanding and Question Answering (arxiv.org)
Sierkinhane/BoxDiff: [ICCV 2023] BoxDiff: Text-to-Image Synthesis with Training-Free Box-Constrained Diffusion (Github)
CTVIS: Consistent Training for Online Video Instance Segmentation (arxiv.org)
The World Outside of AI 🌎
TikTok Wants to Sell Made-in-China Goods to Americans (WSJ)
Education Department opens investigation into Harvard's legacy admissions (AP News)
Heat waves: How does extreme heat affect our physical and mental health? (Vox)
Old houses now cost as much as new houses (Axios)
Study identifies new proteins linked to dementia risk 25 years later (Stat News)
Haptic suits give deaf people a new way to feel live music at Lincoln Center (NPR)
One Last Bite 😋
Will a liberal arts education be more valuable in the age of AI? | Humans of AI Podcast Clip from Evan Mays: From Iron Man and Halo to Hackathon Winner and AI Founder | Humans of AI Podcast #4 - YouTube