Lifeboat Foundation

Meta has created a system that can embed hidden signals, known as watermarks, in AI-generated audio clips, which could help in detecting AI-generated content online.

The tool, called AudioSeal, is the first that can pinpoint which bits of audio in, for example, a full hourlong podcast might have been generated by AI. It could help to tackle the growing problem of misinformation and scams using voice cloning tools, says Hady Elsahar, a research scientist at Meta. Malicious actors have used generative AI to create audio deepfakes of President Joe Biden, and scammers have used deepfakes to blackmail their victims. Watermarks could in theory help social media companies detect and remove unwanted content.

OpenAI has acquired Rockset, which builds tools to drive real-time search and data analytics.

In a post on its official blog, OpenAI said that it would integrate Rockset’s technology to “power [its] infrastructure across products.” Members of Rockset’s team will join OpenAI, and Rockset’s existing customers will be transitioned off of Rockset’s platform “gradually.”

The financial terms weren’t disclosed.

A high school robotics team has built the world’s smallest and cheapest network switch.

The device was created by Murex Robotics and formed by students from Phillips Exeter Academy in New Hampshire.

The invention was necessary because they could not find an affordable embedded ethernet switch for a remotely operated vehicle (ROV) they were building for an underwater drone competition.

Researchers at the University of Manchester have developed an advanced system that enables a robot to achieve record-breaking jumps.

Reconstructing a scene using a single-camera viewpoint is challenging. Researchers have deployed generative artificial intelligence (AI) to achieve this. However, the models can hallucinate objects when determining what is obscured.

An alternate approach is to use shadows in a color image to infer the shape of the hidden object. However, the method falls short when the shadows are hard to see.

To overcome these limitations, the MIT researchers used a single-photon LiDAR. A LiDAR emits pulses of light, and the time it takes for these signals to bounce back to the sensor creates a 3D map of a scene.

Large language models have emerged as a transformative technology and have revolutionized AI with their ability to generate human-like text with seemingly unprecedented fluency and apparent comprehension. Trained on vast datasets of human-generated text, LLMs have unlocked innovations across industries, from content creation and language translation to data analytics and code generation. Recent developments, like OpenAI’s GPT-4o, showcase multimodal capabilities, processing text, vision, and audio inputs in a single neural network.

Despite their potential for driving productivity and enabling new forms of human-machine collaboration, LLMs are still in their nascent stage. They face limitations such as factual inaccuracies, biases inherited from training data, lack of common-sense reasoning, and data privacy concerns. Techniques like retrieval augmented generation aim to ground LLM knowledge and improve accuracy.

To explore these issues, I spoke with Amir Feizpour, CEO and founder of AI Science, an expert-in-the-loop business workflow automation platform. We discussed the transformative impacts, applications, risks, and challenges of LLMs across different sectors, as well as the implications for startups in this space.

As a supplement to optical super-resolution microscopy techniques, computational super-resolution methods have demonstrated remarkable results in alleviating the spatiotemporal imaging trade-off. However, they commonly suffer from low structural fidelity and universality. Therefore, we herein propose a deep-physics-informed sparsity framework designed holistically to synergize the strengths of physical imaging models (image blurring processes), prior knowledge (continuity and sparsity constraints), a back-end optimization algorithm (image deblurring), and deep learning (an unsupervised neural network). Owing to the utilization of a multipronged learning strategy, the trained network can be applied to a variety of imaging modalities and samples to enhance the physical resolution by a factor of at least 1.67 without requiring additional training or parameter tuning.

Researchers have developed a new 3D method that can be used to track fast-moving objects at unprecedented high speeds. The real-time tracking approach, which is based on single-pixel imaging, could be used to improve autonomous driving, industrial inspection and security surveillance systems.

Robots and food have long been distant worlds: Robots are inorganic, bulky, and non-disposable; food is organic, soft, and biodegradable. Yet, research that develops edible robots has progressed recently and promises positive impacts: Robotic food could reduce electronic waste, help deliver nutrition and medicines to people and animals in need, monitor health, and even pave the way to novel gastronomical experiences. But how far are we from having a fully edible robot for lunch or dessert? And what are the challenges?

Scientists from the RoboFood project, based at EPFL, address these and other questions in a new perspective article in the journal Nature Reviews Materials (“Towards edible robots and robotic food”).

“Bringing robots and food together is a fascinating challenge,” says Dario Floreano, director of the Laboratory of Intelligent Systems at EPFL and first author of the article. In 2021, Floreano joined forces with Remko Boom from Wageningen University, The Netherlands, Jonathan Rossiter from the University of Bristol, UK, and Mario Caironi from the Italian Institute of Technology, to launch the project RoboFood.