"A deep neural network model can accurately predict the brain age of healthy patients based on electroencephalogram data recorded during an overnight sleep study." The neural network was accurate to within 4.6 years on average, and identified several things that increase brain age beyond chronological age: epilepsy and seizure disorders, stroke, elevated markers of sleep-disordered breathing, and low sleep efficiency. sleepmeeting.org/artificial-in

"PlayStation is collaborating with Sony's artificial intelligence department in order to create AI 'Agents' that can play games alongside human players." "Sony AI, which we established last year, has begun a collaboration with PlayStation that will make game experiences even richer and more enjoyable. By leveraging reinforcement learning, we are developing Game AI Agents that can be a player's in-game opponent or collaboration partner." ign.com/articles/playstation-g

"The first fiber with digital capabilities, able to sense, store, analyze, and infer activity after being sewn into a shirt." "Until now, electronic fibers have been analog -- carrying a continuous electrical signal -- rather than digital, where discrete bits of information can be encoded and processed in 0s and 1s." "The new fiber was created by placing hundreds of square silicon microscale digital chips into a preform that was then used to create a polymer fiber." news.mit.edu/2021/programmable

Kikodo.io claims to have a Python tutorial system that is "AI powered". I didn't see any details on the website about how the "AI" works. kikodo.io/

Deep learning for data in structured tables. "Tabular data is unique in several ways that have prevented it from benefiting from the impressive success of deep learning in vision and language. First, tabular data often contain heterogeneous features that represent a mixture of continuous, categorical, and ordinal values, and these values can be independent or correlated. Second, there is no inherent positional information in tabular data. arxiv.org/abs/2106.01342

An AI system has found counterexamples for mathematical conjectures. Five of them. Conjectures in graph theory, to be precise. What's interesting is it was done using a reinforcement learning system. The same class of machine learning system that was used to beat Atari games and beat human champions at the Chinese game of Go. arxiv.org/abs/2104.14516

"M4Depth: A motion-based approach for monocular depth estimation on video sequences." The idea is to use a single camera on a drone that is just a regular camera that can't do depth measurements, and use the motion of the drone itself to estimate the depth of everything in the video image. The starting point for their system was a a neural network called PWCNet. PWC is an "optical flow" network. arxiv.org/abs/1709.02371

A light stage is a room-scale (typically) spherical array of brightly-flashing colored lights and cameras and is designed to capture people under any illumination condition. The idea here is to achieve part of what light stages achieve with nothing more than a computer with a bright screen and a webcam in a dark room. You watch Netflix or YouTube and the computer screen sends patterns of light across your face. grail.cs.washington.edu/projec

"The grim reality of jobs in robotics and AI". Crap jobs include: "AI tagging" (annotating images to make training data), "translator" (humans who 'assist' natural language translation systems), "content moderator" (humans who filter abusive, violent, or illegal content, making AI training data in the process, but who suffer psychological trauma), and "warehouse robot" (human who works with real robots in a warehouse). alanwinfield.blogspot.com/2021

"Amazon will extend a ban it enacted last year on the use of its facial recognition for law enforcement purposes." "The web giant's Rekognition service is one of the most powerful facial recognition tools available. Last year, Amazon signed a one-year moratorium that banned its use by police departments." "Amazon has now extended its ban indefinitely." artificialintelligence-news.co

"FICO's State of Responsible AI report shows business leaders are putting little effort into ensuring that the AI systems they use are both fair and safe for widespread use." "Almost 70% of respondents could not explain how specific AI model decisions or predictions are made, and only 35% said their organization made an effort to use AI in a way that was transparent and accountable." zdnet.com/article/fico-report-

"How a largely untested AI algorithm crept into hundreds of hospitals". "The use of algorithms to support clinical decision-making isn't new. But historically, these tools have been put into use only after a rigorous peer review of the raw data and statistical analyses used to develop them. Epic's Deterioration Index, on the other hand, remains proprietary despite its widespread deployment. fastcompany.com/90641343/epic-

The Dark Energy survey cataloging 226 million galaxies (!) has released 3 years of data and 30 research papers. Astrophysicist Becky Smethurst recounts two of the results that got her excited: the dark energy equation of state parameter ("w") that constrains dark energy in the latest cosmological model (the lambda-CDM model), and the "cosmic shear" value ("S8") that represents the distribution of dark matter throughout the universe. youtube.com/watch?v=IHMFDxLcMY

"ReduNet: a white-box deep network from the principle of maximizing rate reduction." Basically what they are claiming to do is to design the architecture of a neural network, but do so "from first principles" based on mathematics and not trial-and-error. arxiv.org/abs/2105.10446

"AI-powered drone deployed in Libya has possibly killed people without any human interference." "The drone, Kargu-2, is made by a Turkish company (STM) and fitted with a payload that explodes once it makes an impact or is in close proximity with its AI-identified target. It is not clear whether the attacks resulted in any deaths." iafrikan.com/2021/06/02/artifi

"Chinese AI lab challenges Google, OpenAI with a model of 1.75 trillion parameters." "Wudao has 150 billion more parameters than Google's Switch Transformers, and is 10 times that of OpenAI's GPT-3, which is widely regarded as the best model in terms of language generation." "Unlike conventional deep learning models that are usually task-specific, Wudao is a multi-modal model trained to tackle both text and image, two dramatically different sets of problems." en.pingwest.com/a/8693

"A piece of space debris too small to be tracked has hit and damaged part of the International Space Station - namely, the Canadarm2 robotic arm." sciencealert.com/space-debris-

Photos from China's Mars rover (Zhurong) from the website of China's National Space Administration (CNSA). cnsa.gov.cn/n6759533/c6812021/

"Surviving an in-flight anomaly: what happened on Ingenuity’s sixth flight." "Approximately 54 seconds into the flight, a glitch occurred in the pipeline of images being delivered by the navigation camera. This glitch caused a single image to be lost, but more importantly, it resulted in all later navigation images being delivered with inaccurate timestamps." mars.nasa.gov/technology/helic

"Adding noise -- also known as pixilation -- along multiple layers of a network provides a more robust representation of an image that's recognized by the AI and creates more robust explanations for AI decisions." "It's about injecting noise into every layer. The network is now forced to learn a more robust representation of the input in all of its internal layers." utsa.edu/today/2021/05/story/j

Show older
Mastodon

Server run by the main developers of the project 🐘 It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!