🥟 Chao-Down #123 US Military is testing out Generative AI, Google's medical AI chatbot deployed to hospitals, Wimbledon considers replacing line judges with AI

In the aftermath of ChatGPT's launch, there have been notable concerns about the rise of plagiarism among students. To address these fears, several startups began developing products claiming to identify whether text was written by a human or a machine.

However, recent research indicates that these tools can be easily deceived. Debora Weber-Wulff, a professor of media and computing at HTW Berlin, collaborated with researchers to evaluate the performance of 14 detection tools, including Turnitin, GPT Zero, and Compilatio, in detecting text generated by ChatGPT. The study revealed that these tools struggled to identify ChatGPT-generated text when it was slightly altered by humans and obscured using paraphrasing tools.

By making minor adaptations to the AI-generated essays, students could easily evade detection. We still have a lot more work to do.

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

EU, Japan look to partner on A.I., chips amid China 'de-risking' push (CNBC)

Inside the AI Factory: the humans that make tech seem human (The Verge)

The US Military Is Taking Generative AI Out for a Spin (Bloomberg)

Full-Body AI Scans Could Be the Future of Preventive Medicine (CNET)

Google’s medical AI chatbot is already being tested in hospitals (The Verge)

Alibaba launches A.I. tool to generate images from text (CNBC)

In – or out? Wimbledon considers replacing line judges with AI (The Guardian)

AI-text detection tools are really easy to fool (MIT Technology Review)

Always be Learnin’ 📕 📖

A Survey on Evaluation of Large Language Models (arxiv)

What AI can do with a toolbox... Getting started with Code Interpreter (oneusefulthing.org)

Projects to Keep an Eye On 🛠

PEP 703 – Making the Global Interpreter Lock Optional in CPython (peps.python.org)

kudoai/chatgpt.js: 🤖 A powerful, open source client-side JavaScript library for ChatGPT (Github)

CStanKonrad/long_llama: LongLLaMA is a large language model capable of handling long contexts. It is based on OpenLLaMA and fine-tuned with the Focused Transformer (FoT) method. (Github)

The Latest in AI Research 💡

Internet-augmented language models through few-shot prompting for open-domain question answering (arxiv)

Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks (arxiv)

Mixture-of-Experts Meets Instruction Tuning: A Winning Combination for Large Language Models (arxiv)

Inverse Reinforcement Learning without Reinforcement Learning (gokul.dev)

The World Outside of AI 🌎

China Controls Minerals That Run the World—and It Just Fired a Warning Shot at U.S. (WSJ)

Threads is now the fastest platform to reach one million active users (lifestyleasia.com)

Twitter warns it could sue Meta over ‘copycat’ Threads app (The Verge)

Tips for coping with increasingly hot weather (The Washington Post)

Drowning Is No. 1 Killer of Young Children. U.S. Efforts to Fix It Are Lagging. (The New York Times)

One Last Bite 😋

Check out this clip from the Humans of AI podcast!

From “Cassandra Ansara: Safeguarding against Prompt Injection | Humans of AI Podcast #1”