Attack on AI Chatbots, AI Lies, and more! | Fetch.ai Newsletter | Issue 17/08/2023

Welcome to the first edition of the Fetch.ai Newsletter!?

Delivered to your inbox every fortnight, we'll bring you the latest insights and innovations in the world of artificial intelligence. This fortnightly digest is your essential guide to all things AI. Join us on this exciting journey through the ever-evolving landscape of artificial intelligence.

In This Issue

• A new attack impacts major AI chatbots.
• What can you do when AI lies about you?
• This new tool could protect your pictures from AI manipulation.
• The Big Story: Can There Be Cooperation on AI Regulations?
• Quick Scoops: OpenAI bankruptcy speculation, IBM x NASA x AI.

Must Read

A new attack impacts major AI chatbots.

Carnegie Mellon University researchers have discovered a startling vulnerability in popular AI chatbots like ChatGPT. By using specially crafted text strings that appear as gibberish to humans, they can manipulate chatbots into revealing prohibited information. The problem, according to CMU's associate professor Zico Kolter, represents a fundamental weakness, not a simple bug. While efforts to block these attacks have begun, a complete solution remains elusive. This revelation raises urgent questions about AI security. Read more to know how this could impact our technology-dependent world.?

What Can You Do When A.I. Lies About You?

Marietje Schaake, a prominent Dutch politician and director at Stanford University's Cyber Policy Center, found herself falsely labeled a "terrorist" by Meta's AI chatbot, BlenderBot 3. This alarming incident underscores a broader concern with AI's struggles in accuracy, including the creation of fabricated information that can potentially ruin reputations. Efforts are underway to improve AI's reliability, but with persistent challenges, questions arise about the technology's safety and ethical implications. Are these mistakes mere hiccups or indicative of a more significant problem in AI? ?Full write-up by NYT here .

This new tool could protect your pictures from AI manipulation

The ease of image manipulation through AI has led to growing concerns about deepfakes and potential abuse. New tools, PhotoGuard and Glaze, are being developed to protect images by altering them in subtle ways that thwart AI systems. PhotoGuard safeguards personal photos, while Glaze protects artists' copyrighted styles. While promising, these tools require widespread adoption by tech companies to be truly effective. The situation calls for a deeper examination of image security in the era of advanced AI. Read more about the tools and the broader implications of AI-driven image manipulation .

The Big Story: Can There Be Cooperation on AI Regulations?

.Today's big story explores the lightning-fast advancements in AI and the increasing calls from industry experts for international regulations. Concerns encompass fairness, bias, potential disruption, and societal changes. While regulations are emerging, they remain fragmented across global powers. The need for a cohesive international approach, possibly through a Global Digital Compact or global summit, is underscored to ensure that the development of AI is balanced between caution and unlocking its vast potential benefits.

Quick Scoops

Important AI news scoops that you need to know:

?? ChatGPT’s soaring costs may lead to OpenAI’s bankruptcy

?? AI helps quadriplegic men move and feel again

?? IBM and NASA tackle climate change using AI

?? AI cameras help track wildlife in the UK

?? Meta’s new tool for creating music from text is here

We hope you enjoyed our first edition of the Fetch.ai Newsletter! If you want to read more of these stories - be sure to subscribe to our newsletter to stay up-to-date on the latest AI news.