Rabbit's deadly data leak • AI creates simulated internet • Stability AI stays alive • AI news this week
Welcome to the AI Search newsletter. Here are the top highlights in AI this week.
Rabbit's Critical API Key Leak Exposes User Data and R1 Devices
The Rabbitude team discovered that Rabbit Inc. has been negligent in securing their API keys, leaving critical hardcoded keys accessible in their codebase. These keys grant unauthorized access to sensitive information, including all R1 responses ever given, which may contain personal information. The most concerning key is for ElevenLabs, allowing full privileges to manipulate R1 devices, including changing voices, deleting voices, and even crashing the RabbitOS backend.
WebSim: This AI Might Be the Future of Internet
WebSim is a revolutionary platform that allows users to create and simulate web applications by just prompting. Think of this as using an AI to create an alternative internet. See a full breakdown here:
Stability AI Secures Funding from Sean Parker and Greycroft to Stay Afloat
Stability AI, the company behind the popular open-source image-generating model Stable Diffusion, has received a financial lifeline from investors including Sean Parker, Greycroft, and others. The exact amount of funding remains undisclosed, but it's reportedly enough to help the company recover from a cash crunch and unpaid cloud bills. Stability AI's co-founder and former CEO, Emad Mostaque, was ousted in March following mismanagement allegations, and the company has since pivoted to a subscription model and explored other revenue streams.
ChatLLM by Abacus
ChatLLM by Abacus AI is an integrated platform that allows enterprises to use multiple LLMs, deploy custom agents, and collaborate with team members. Choose from state-of-the-art LLMs such as GPT-4o, Claude 3 Opus, and their new open source Smaug.
Hallo: A Free Open-Source AI for Animating Faces
This research presents a breakthrough in portrait image animation driven by speech audio input, achieving highly realistic and dynamic results. The innovative approach abandons traditional parametric models, instead employing an end-to-end diffusion paradigm and a hierarchical audio-driven visual synthesis module to precisely align audio inputs with visual outputs. Comprehensive evaluation demonstrates significant improvements in image and video quality, lip synchronization precision, and motion diversity, paving the way for more effective personalization in portrait animation. Here’s the full breakdown:
Breakthrough in Robotic Hand Dexterity with Tactile Fingertips
Researchers have achieved a significant breakthrough in robotic hand dexterity by creating a four-fingered robotic hand with artificial tactile fingertips. This innovative hand can rotate objects in any direction and orientation, even when upside down, thanks to its advanced tactile sensors that mimic human skin. The robotic hand's ability to sense touch and adjust its grip has enabled it to perform complex tasks with precision, paving the way for potential applications in industries such as manufacturing, healthcare, and recycling.
ALOGIC
ALOGIC offers a wide range of professional home/office hardware, including their Clarity series of professional 4K touchscreen monitors, wireless chargers, docking stations, and more. Enhance your productivity and elevate your digital lifestyle with ALOGIC!
Sony Unveils AI for Generating High-Quality Instrumental Accompaniments in Music Production
Sony has introduced an innovative AI system, called Diff-A-Riff, that can generate high-quality instrumental accompaniments for any music track. This AI-powered tool uses latent diffusion models and consistency autoencoders to produce accompaniments that seamlessly match the style and tonality of a given musical context. With its versatile control, users can condition both audio and text prompts, and it can generate accompaniments for any instrument type. The system has been tested and has produced high-quality outputs that are indistinguishable from those played by human musicians.
New AI-powered tool can detect fake videos with high accuracy
Researchers have developed a new tool called DIVID (DIffusion-generated VIdeo Detector) that can accurately detect AI-generated videos with a 93.7% success rate. The tool works by reconstructing a video and analyzing the differences between the original and reconstructed videos, exploiting the fact that AI-generated videos tend to have more "statistical means" content, such as pixel intensity distributions and texture patterns. The researchers believe that DIVID has the potential to stop scammers who use AI-generated videos and protect society from deepfakes.