In order to find the country location for each video in previous Y ouTube-based datasets (e.g., Kinetics, HACS, etc.), we used the public Y ouTube API. 'The geolocation information associated with the video. In our measure, roughly 8% of the videos had such geolocation. We then used reverse-geocode library https://pypi.org/project/reverse-geocode/

AI is the hot buzzword in tech. We're hearing it repeatedly from major tech companies, as they integrate AI models into their products. It's a pair of announcements in recent weeks, with the most current an update to Google's Enhanced Browsing protection, which screens your browsing data in real time for threats like malware, fraudulent websites, and sketchy extensions. The opt-in program received an official boost from AI, which Google says aids the detection of internet dangers. Details have yet to be shared of how exactly the AI screening works, which mirrors another piece of news that involves Chrome and its password manager.

While large models pre-trained on high-quality data exhibit excellent performance across various reasoning tasks, including mathematical reasoning (e.g. GSM8k, MultiArith), specializing smaller models to excel at mathematical reasoning remains a challenging problem. Common approaches to address this challenge include knowledge distillation, where smaller student models learn from large pre-trained teacher models, and data augmentation, such as rephrasing questions. Despite these efforts, smaller models struggle with arithmetic computations, leading to errors in mathematical reasoning. In this work, we focus on leveraging a programmatically generated arithmetic dataset to enhance the reasoning capabilities of smaller models. We investigate two key approaches to incorporate this dataset -- (1) intermediate fine-tuning, where a model is fine-tuned on the arithmetic dataset before being trained on a reasoning dataset, and (2) integrating the arithmetic dataset into the instruction-tuning mixture, allowing the model to learn arithmetic skills alongside general instruction-following abilities. Our experiments on multiple reasoning benchmarks demonstrate that incorporating an arithmetic dataset, whether through targeted fine-tuning or within the instruction-tuning mixture, enhances the models' arithmetic capabilities, which in turn improves their mathematical reasoning performance.

In this paper, we build upon two major recent developments in the field, Diffusion Policies for visuomotor manipulation and large pre-trained multimodal foundational models to obtain a robotic skill learning system. The system can obtain new skills via the behavioral cloning approach of visuomotor diffusion policies given teleoperated demonstrations. Foundational models are being used to perform skill selection given the user's prompt in natural language. Before executing a skill the foundational model performs a precondition check given an observation of the workspace. We compare the performance of different foundational models to this end as well as give a detailed experimental evaluation of the skills taught by the user in simulation and the real world. Finally, we showcase the combined system on a challenging food serving scenario in the real world. Videos of all experimental executions, as well as the process of teaching new skills in simulation and the real world, are available on the project's website.

Getting the'side-eye' look from your dog can certainly make it seem like they are judging you, and a new study suggests they really could be. Researchers at the University of Vienna found that pooches can tell when we dangle a treat out of their reach to be cruel, or when it is just an accident. What's more, they also act differently towards us depending on our perceived intention, as they appear more patient with the clumsy than the mean. Only a few animals have previously been shown to be able to make social evaluations of humans in this manner, including chimpanzees, capuchin monkeys and African grey parrots. The team recruited 96 pet dogs for the experiment, and each was presented with one of two scenarios.

Pet dogs know when you intend to give them a treat, even if you drop it where they can't get to it Dogs can understand when humans mean well, even if they don't get what they want from us. Prior to this work, the ability to distinguish between a human being unwilling or unable to perform a task had only been found in non-human primates. The close social bond between humans and canines is well established, but researchers have a limited understanding of if and how dogs comprehend human intent. To see if pet dogs can distinguish between intentional and accidental actions by strangers, Christoph Völter at the University of Veterinary Medicine Vienna in Austria and his colleagues ran tests with humans offering dogs food while the animals' body movements were tracked using eight cameras. Each dog and human were separated by a transparent plastic panel with holes that a slice of sausage could be passed through.

Astrophysicians have used AI to discover 250 new stars in the Milky Way, which they believe were born outside the galaxy. Caltech researcher Lina Necib named the collection Nyx, after the Greek goddess of the night. She suspects the stars are remnants of a dwarf galaxy that merged with the Milky Way many moons ago. To develop the AI, Necib and her team first tracked stars across a simulated galaxy created by the Feedback in Realistic Environments (FIRE) project. They labeled the stars as either born in the host galaxy, or formed through galaxy mergers.

Moments after plugging in Amazon's new Alexa-connected microwave I was about to review, the first thing I noticed was the word that popped up on its screen: FAIL. It hadn't--it had just jumped the gun a bit in the connection process, but that word hung over the testing process in surprising ways. The AmazonBasics Microwave is a 700-watt, 0.7-cubic foot appliance that costs a mere 60 dollars and connects with Amazon devices like the Echo, allowing you to control many of the microwave's functions with your voice. Say "Alexa, microwave 30 seconds" and the appliance starts whirring away on high. Trying "Alexa defrost ten ounces of fish" will result in more-regulated microwave blasts.

In his biting, much-cheered defense of the work of local newspapers on his show "Last Week Tonight," the red-hot HBO satirist John Oliver had much fun at the expense of the role of "artificial intelligence" in modern journalism. Oliver's highly entertaining piece -- which quickly garnered well in excess of 4 million views -- contrasted the current enthusiasm of some publishing executives, including the ones who currently pay my salary, for various automated manifestations of reporting, editing and news distribution with what you might call the old-fashioned, sentimental view of the profession: the notepad-wielding reporter at the quotidian school board meeting, fighting corrupt politicians and delivering the truth to your stoop. There is a lot to unpack in Oliver's 19-minute segment and various levels of irony at work. For starters, there's this: In decrying the tendency of panicked newspapers to veer toward populist click-bait, Oliver cleverly created, well, his own populist click-bait. Oliver humbly and openly acknowledged how much his show depends on newspapers for its material -- thank you very much on behalf of my hard-working colleagues.