• Chaos Theory
  • Posts
  • 🥟 Chao-Down #193 US Space Force pauses use of AI tools over data security risks, Baidu claims its AI is as good as ChatGPT, President Biden cuts China and other countries off from more NVIDIA chips

🥟 Chao-Down #193 US Space Force pauses use of AI tools over data security risks, Baidu claims its AI is as good as ChatGPT, President Biden cuts China and other countries off from more NVIDIA chips

Plus, Microsoft launches an AI bounty program for security researchers to discover vulnerabilities in Bing AI

Can AI chatbots like ChatGPT guess your personal information as you talk with it? Recent research seems to suggest so.

The study, conducted by researchers at ETH Zurich, tested language models from OpenAI, Google, Meta, and Anthropic, all of which showed the capability to infer private data including people’s race, location, and occupation.

Because large language models are trained on vast amounts of web content including census data, many of which can contain personally identifiable information, it makes the challenge of anonymizing interactions with AI especially difficult to address.

Scammers could potentially exploit this ability to extract sensitive data from unsuspecting users, while companies could use it for targeted advertising.

Have we entered a world where our data is just one ChatGPT-conversation away from being leaked? Will we even care?

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

AI Chatbots Can Guess Your Personal Information From What You Type (WIRED)

Baidu Says Its AI as Good as ChatGPT in Big Claim for China (Bloomberg)

US Space Force pauses use of AI tools like ChatGPT over data security risks (Reuters)

United States Space Force präsentiert Logo – Design Tagebuch

Biden to cut China off from more Nvidia chips, expand curbs to more countries (Reuters)

Microsoft AI Bounty - The Microsoft AI bounty program invites security researchers from across the globe to discover vulnerabilities in the new, innovative, AI-powered Bing experience (MSRC)

Generative AI deals are taking a dive as investors pick their spots (qz.com)

Always be Learnin’ 📕 📖

How to lose at Generative AI! - by Sangeet Paul Choudary (Substack)

Commercializing AI in Healthcare: The Enterprise Buyer Perspective (Andreessen Horowitz)

Failing, FTW!: Why you should fail 50% of your experiments (Substack)

Projects to Keep an Eye On 🛠

microsoft/prompts-for-edu: Prompts for Education: Enhancing Productivity & Learning (Github)

michaelfeil/infinity: Infinity is a high-throughput, low-latency REST API for serving vector embeddings, supporting a wide range of sentence-transformer models and frameworks. (Github)

voxel51/fiftyone: The open-source tool for building high-quality datasets and computer vision models (Github)

The Latest in AI Research 💡

InfoCL: Alleviating Catastrophic Forgetting in Continual Text Classification from An Information Theoretic Perspective (arxiv)

Srijith-rkr/Whispering-LLaMA: EMNLP 23 - Integrating Whisper Encoder to LLaMA Decoder for Generative ASR Error Correction (Github)

UniSim: Learning Interactive Real-World Simulators (universal-simulator.github.io)

The World Outside of AI 🌎

Half Of All Skills Will Be Outdated Within Two Years, Study Suggests (Forbes)

The Curse of the Creator Economy (WIRED)

Renting can age you faster than smoking, obesity: study (NYPost)

Kids Suing Social Media Over Addiction Find a Win Amid Losses (Bloomberg)

Popular baby names: The psychology of growing up with an extremely common name (qz.com)

One Last Bite 😋