A local cathedral is remembering a devout, dedicated nun following her death after a wreck on Thursday. A local cathedral is remembering a devout, dedicated nun following her death after a wreck on Thursday. A local cathedral is remembering a devout, dedicated nun following her death after a wreck on Thursday. A local cathedral is remembering a devout, dedicated nun following her death after a wreck on Thursday. Get the latest news stories of interest by clicking here. “Shock firstly, and sadness. She was an incredibly warm and generous part of this community; it’s really difficult to imagine life around here without Sister Veronica,” said Nolan Reilly, director of music for the Cathedral of Our Lady of Perpetual Help. The pews inside the Cathedral of Our Lady of Perpetual Help will be met with silence this week as they mourn the loss of Sister Veronica Higgins. In video shared by the cathedral, Higgins can be seen doing what she loved for nearly 30 years, singing with the church’s choir. The cathedral shared photos and videos of Higgins as they remember the legacy she left behind. “She was a very gentle and caring person, but she was also a very powerful person who advocated for everyone she loved,” Reilly said. >> Download the KOCO 5 App Higgins died after a medical episode caused her to veer off the roadway, and now, she’s being described as a giant in the community. “She was a very prominent figure, who was beloved, so a lot of people are grieving right now,” Reilly said. A part of that was the way she moved others through her music inside the cathedral. “She used it to relate to other people and used it as a tool to evangelize, and above all, she was one of the most faithful people I’ve ever met. And she used it to communicate with God,” Reilly said. And now, as they prepare to lay Sister Veronica Higgins to rest, Reilly wants people to know she was a champion for education, children, the poor and the church. He said she was a joyful woman who will be missed by many. Top Headlines Hearst Television participates in various affiliate marketing programs, which means we may get paid commissions on editorially chosen products purchased through our links to retailer sites.
Fireworks News Volunteers visit with dogs at Dallas Animal Services. An innovative program at Dallas' animal shelter turned out to be a major hit for animals and people alike. Called Calming the Canines, it was an event that took place on New Year's Eve at Dallas Animal Services' facility on 1818 N. Westmoreland Rd. to help reduce stress for cats, dogs, and other animals at the shelter during fireworks. Nearly 300 people turned out for the event, which involved sitting outside the kennel in which the animals were confined and engaging them in activities such as reading or singing. Many brought toys and food donations, and a few adopted animals and signed up for the shelter's foster program.
With 307 public kennels at the facility, it resulted in just about every animal receiving some kind of attention. While the idea of reading to animals at shelters may sound unusual, it's not a new thing, and studies have shown it can have a positive effect. Other animal organizations around Dallas have hosted similar reading nights — but none on New Year's Eve. The timing was to address the impact that fireworks can have on animals, even those confined inside shelter walls. Cats, dogs, birds, and wildlife have better hearing than humans, and are more affected by fireworks, which come off as a threat. The shelter ordinarily closes at 7 pm but they opened the doors from 10:30 pm to 12:30 am when fireworks would be at their peak. One notable aspect at this event was the wide range of people who showed up, not just from Dallas but outlying cities like Plano and McKinney — everyone from a big family with a bunch of kids, to a mom with her son who was a Marine, to five National Honors Society students who needed to fulfill volunteer hours.
Many were visiting the shelter for the first time. And everyone followed instructions — which is to say that they came armed with books, not to mention quilts, blankets, bags of treats, and other donations. Some went specifically to the library to find books to bring. One man who came was a former dog owner who'd just moved from Houston and wasn't able to have a pet at his place, but said he just wanted to help the pets get through the night. One woman played a harp. There were also generous company donations including pizzas donated by Dallas pizza chain Cane Rosso, plus beverages and manpower from Rahr Brewing. Unfortunately, an event like this does not solve the problem which is an epidemic of illegal fireworks being set off in areas such as South Dallas, where loose animals are already a problem and where residents do not properly care for their pets; and police departments across North Texas that are already grappling with other holiday-related issues. In the days after fireworks-heavy nights like New Year's Eve and July 4th, most shelters witness an increased intake of lost and loose animals, but that doesn't even address the cats and dogs who've bolted during fireworks in fear, found dead in the ensuing days.
But on this one night, 300 or so people at least helped make a difference at the shelter itself. "Definitely a big part of the story is how the community came together," said one DAS volunteer. Animal News With sub-freezing temperatures predicted for North Texas in the next few days, the SPCA of Texas is recommending that pet owners protect their pets from harsh outdoor conditions. Keep your pet indoors Freezing temperatures can be dangerous and even deadly for companion animals, so it is best to keep your pet indoors as much as possible. A good rule of thumb is that if you are cold outside, your pet will be, too. Wet and cold weather can lead to hypothermia or pneumonia in animals. Be especially cautious with very young or very old animals, because they are more susceptible to suffering medical issues due to the cold.
Limit outside time to quick walks or bathroom breaks, and consider providing your pet with a pet sweater and booties to protect their paws from ice and/or snow. Avoid pavement and walk on grass when possible. Outdoor living quarters The SPCA of Texas never recommends leaving pets outside full time. And it is against the law in the state of Texas to leave your pet outdoors in extreme temperatures without appropriate shelter. However, if you must subject your pets to being outside for extended periods of time, weatherproof their living quarters. That means: Food & drink Monitor the time your pets spend outdoors and be sure they always have fresh water that is not frozen, in a non-metal bowl to drink. Outdoor dogs need more calories in the winter to produce body heat, so increase the amount you feed your pets if they stay outdoors for long periods of time.
Things not to do Some pet owners assume their pets can withstand the cold weather because they have fur. But this is not always the case. Indoor dogs shed their undercoats and should never be made to stay outside for extended periods of time. Never shave your pets down to the skin in the winter; leave their coats long for more warmth. When you bathe your pets, completely dry their coats before letting them go outdoors. Things to watch out for Be diligent regarding chemicals and pets in vehicles, as follows:
Successive DDoS attacks and a data breach force the Internet Archive offline. Meanwhile, users on social media are blasting the hacker who has claimed responsibility. The Internet Archive could be offline for a while to prioritize data security after a hacker breached the nonprofit and stole data on 31 million users. In addition to the data breach, the organization has been fending off repeated DDoS attacks since Tuesday. On Thursday, the Internet Archive’s site was briefly restored, but then another DDoS attack knocked the domain and its Wayback Machine offline again. The repeated attacks prompted the archive’s founder, Brewster Kahle, to tweet on Thursday morning that the Internet Archive “is being cautious and prioritizing keeping data safe at the expense of service availability. Will share more as we know it.” The hacker made the breach known after briefly defacing the Internet Archive the day before with a pop-up that mentioned the nonprofit had suffered “a catastrophic security breach.” Kahle has since confirmed the breach, saying usernames, email accounts, and hashed passwords were stolen from the site. The breach affects 31 million user accounts, according to data breach notification site Have I Been Pwned, which received a copy of the stolen data from the alleged attacker. Users can find out if they’re affected by using the HIBP site. It’s unclear how the data was stolen. But according to Kahle, the hacker defaced the Internet Archive by hijacking a JavaScript library for one of its sites. In the meantime, the Twitter/X account @Sn_darkmeta has claimed responsibility for the DDoS attacks. “They are under attack because the archive belongs to the USA, and as we all know, this horrendous and hypocritical government supports the genocide that is being carried out by the terrorist state of ‘Israel,’” the account posted. But the justification is facing blowback from users since the Internet Archive is merely a San Francisco-based nonprofit devoted to acting as a free online library. As a result, some users are condemning @Sn_darkmeta for carrying out the DDoS attacks. In a since-deleted tweet, @Sn_darkmeta responded by saying: “We are not interested in your dog barking behind a mobile screen. If the Internet Archive was shut down for all countries and users, it’s only a taste to experience deprivation. You’re protesting and crying just because you can’t enjoy a free service.” “Imagine the people in Sudan and Gaza, millions are being subjected to genocide and you’re just a bunch of fools reading about these events,” the post added. “Innocent people don’t need documentation because the truth is already clear: America, Europe and Israel are the greatest cancers in this universe.” The account has since published a video in the Russian language describing itself as hacktivist group. Sign up for SecurityWatch newsletter for our top privacy and security stories delivered right to your inbox.
A British national was among at least 14 people killed in the vehicle attack in New Orleans on New Year's Day, the Foreign Office has confirmed. In a statement, the Foreign Commonwealth and Development Office (FCDO) said it was supporting the deceased person's family. An FCDO spokesperson said: "We are supporting the family of a British National who has died in New Orleans and are in contact with local authorities." During the attack, a man in a pickup truck ploughed through crowds on the city's Bourbon Street before being killed by police. This breaking news story is being updated and more details will be published shortly. Please refresh the page for the fullest version. You can receive Breaking News on a smartphone or tablet via the BBC News App. You can also follow @BBCBreaking on X to get the latest alerts. Trump had attempted to use his presidential election victory to dismiss the case against him. Similarities between the two incidents appear coincidental, with no evidence the alleged perpetrators knew each other. Republican Mike Johnson could only lose two votes from his own party in his bid for speaker re-election, and he nearly did. Friday's vote underscores the challenges for Trump in keeping House Republicans united to legislate his agenda. The suspect is seen placing a device near the Democratic National Committee headquarters a night before the Capitol riot. Copyright 2025 BBC. All rights reserved. The BBC is not responsible for the content of external sites.Read about our approach to external linking.
Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript. Advertisement Light: Science & Applicationsvolume 14, Article number: 34 (2025) Cite this article Metrics details Current trends in artificial intelligence toward larger models demand a rethinking of both hardware and algorithms. Photonics-based systems offer high-speed, energy-efficient computing units, provided algorithms are designed to exploit photonics’ unique strengths. The recent implementation of cellular automata in photonics demonstrates how a few local interactions can achieve high throughput and precision. Current artificial intelligence (AI) models based on neural networks are gaining previously inaccessible cognitive and creative abilities with the continuous increase in their scale. State-of-the-art models now tend to double their sizes every year, as shown in Fig. 1a, reaching trillions of parameters today. In addition to better performances in their training tasks, as the models are scaled up, they have also been observed to start performing new tasks that they were not trained for1. Fig. 1 illustrates this phenomenon, showing language models obtain capabilities outside of their training after reaching a certain level of complexity. This expanded skill set, coupled with wider adoption across various sectors, is driving a rapid increase in global computing resource and energy demands for AI, currently doubling every 100 days2. The corresponding environmental impact of this energy-hungry technology necessitates the development of more compact AI models and more efficient hardware, while maintaining high performance. a The trend of the total number of parameters of the state-of-the-art AI models over time, each data point refers to such a model (Epoch (2024) – with major processing by Our World in Data). b–d Different examples of emergent capabilities in large-scale language models. As the scale of these models trained on generic language datasets increases, they become able to perform tasks beyond those for which they are explicitly trained. b Accuracy on arithmetic operations task17. c Translation accuracy between International Phonetic Alphabet and English17d Accuracy on multitask language understanding, a benchmark containing 57 tasks, ranging from computer science to law18 Different machine learning methods address the goal of achieving competitive accuracies with smaller and lighter models. As one of the earlier techniques, pruning reduces the size of neural networks by determining less important connections after training and eliminating them3. Knowledge distillation trains a smaller model with the intermediate activations of a larger model, achieving similar performance with fewer parameters4. The method called quantization, which is simply decreasing the bit depth of model parameters and/or activations during inference, for instance from 16 bits to 8 bits, also resulted in larger throughput with the same computational resources5. Relying on randomly initialized, fixed hidden layers that do not require gradient-based training, Extreme Learning Machines (ELM)6 and reservoir computing7 decrease the number of trainable parameters. Another advantage of these architectures is the possibility of low-power, high-dimensional and parametric physical events to perform their fixed layers with high efficiency. Alongside advances in AI algorithms, the use of alternative modalities for hardware holds the potential to reduce the environmental impact of this technology. Photonics is one of the promising candidates since it can sustain larger bandwidths and lower losses compared to digital electronics. Mature photonic technologies, such as integrated and spatial light modulators, enable the implementation of various AI models, including fully programmable architectures8,9 and configurations with fixed layers, whose functionality comes from physical interactions such as multimode lasing10, nonlinear frequency conversion11 or random scattering12. Besides power efficiency, another advantage of high-dimensional nonlinear physical events is their suitability for computing complex tasks with a minimal number of parameters13. This advantage has been demonstrated with spatiotemporal nonlinearities in multimode fibers, the selection from a large set of readily available connectivities achieved the accuracy of artificial neural networks with over two orders of magnitude more parameters than the optical implementation14. Compared to global connections in layers such as fully connected and attention, processing information with local connections in an AI model results in more compact architectures, one very popular and influential example being convolutional layers. Neural cellular automata (NCA), inspired by traditional cellular automata in which each cell of the system evolves according to local rules that depend on neighboring cell states, use differentiable, continuous-valued functions to define these interactions15. This design allows NCA to perform complex tasks through simple update rules. The “neural” or differentiable nature of NCA enables the definition of a downstream task for the local interactions and subsequent training of interaction weights accordingly. In the study by Li et. al. from the California Institute of Technology, the downstream task was defined as the classification of the overall pattern formed by pixels (or “cells”, in the context of cellular automata), and a photonic system has achieved the implementation of the NCA16. The computational model depending on the recurrent updates to the individual cell values according to the interaction rules was proved to be a convenient match with the capabilities of photonics. As shown in Fig. 2, the various computational functionalities required by the algorithm were realized by different optical components. During inference, the fixed interactions between cells were implemented with a variable optical attenuator, while second harmonic generation in the periodically poled lithium niobate acts as the nonlinear activation function. The updated cell values were then detected and returned to the optical domain through a high-speed electro-optic modulator. a Working principle of neural cellular automata. Each pixel/cell interacts with its neighboring cells with a set of weights, trained with gradient descent. The final values of these cells represent an individual local decision about the global distribution. b The local interaction scheme behaves as a perceptron, whose output becomes the value of the cell in the next step. While the weighted sum is performed in photonics by the combination of the outputs of variable optical attenuators, c the pump depletion in a periodically poled lithium niobate waveguide, d serves as the nonlinear activation Leveraging the immense data rate of the modulator, the optoelectronic system achieved predictions at a state-of-the-art rate of 1.3 μs per frame. This high throughput was further enabled by the simplicity of the local interaction model, that was defined by only 3 parameters, allowing each cell to compute its next state based on its current state and the states of its two neighbors. For the final binary classification, a majority “vote” was conducted across all cells, with classification as “1” if the majority of cells exceeded a threshold value and “0” otherwise. The classification precision reached 98.0%, closely matching the ideal simulation accuracy of 99.4%, due to the proposed mixture of experts approach’s resilience to experimental nonidealities, such as noise and device imperfections. A remarkable finding of the paper by Li, et al., is that good accuracy can be obtained in the classification of images for the MNIST fashion database with 2 classes, In order to understand whether this is due to the specifics of the NCA architecture used, we implemented on the same database a more familiar multilayer network consisting of a single convolutional layer with a 2-by-2 kernel followed by a similar output classification layer. With a total of 7 parameters, this network achieved a similar 98.3% test accuracy while processing an image in 18.6 μs (instead of 1.3 μs) with a batch size of 1024, on an NVIDIA T4 GPU. We conclude, therefore, a strength of the photonic approach is that even compared to the highly optimized and parallelized GPU hardware, it was able to operate at a higher speed. This photonic implementation of neural cellular automata (NCA) illustrates how photonics could address the explosion of model sizes and the environmental footprint of AI by utilizing high-speed hardware and physical interactions as computing units. Given the development of algorithms tailored to these platforms—considering the unique advantages and limitations of photonics rather than those of general-purpose digital hardware—photonics may offer a compelling solution. As demonstrated here, aligning the algorithm’s requirements with photonic capabilities enables implementations with high precision and throughput that could contribute to the scaling of AI sustainably. Wei, J. et al. Emergent abilities of large language models. Trans. Mach. Learn. Res.2022 (2022). Zhu, S. Q. et al. Intelligent computing: the latest advances, challenges, and future. Intell. Comput.2, 0006 (2023). Article Google Scholar Sietsma & Dow. Neural net pruning-why and how. Proceedings of the IEEE 1988 International Conference on Neural Networks. San Diego, CA, USA: IEEE, 1988, 325-333, https://doi.org/10.1109/ICNN.1988.23864. Hinton, G., Vinyals, O. & Dean, J. Distilling the knowledge in a neural network. Print at https://doi.org/10.48550/arXiv.1503.02531 (2015). Grattafiori, A. et al. The llama 3 herd of models. Print at https://doi.org/10.48550/arXiv.2407.21783 (2024). Huang, G. B., Zhu, Q. Y. & Siew, C. K. Extreme learning machine: theory and applications. Neurocomputing70, 489–501 (2006). ArticleMATH Google Scholar Schrauwen, B., Verstraeten, D. & Van Campenhout, J. M. An overview of reservoir computing: theory, applications and implementations. Proceedings of the 15th European Symposium on Artificial Neural Networks. Bruges, Belgium: ESANN, 2007, 471-482. Lin, Z. J. et al. 120 GOPS Photonic tensor core in thin-film lithium niobate for inference and in situ training. Nat. Commun.15, 9081 (2024). ArticleMATH Google Scholar Li, J. X. et al. Massively parallel universal linear transformations using a wavelength-multiplexed diffractive optical network. Adv. Photonics5, 016003 (2023). ArticleADS Google Scholar Skalli, A. et al. Photonic neuromorphic computing using vertical cavity semiconductor lasers. Opt. Mater. Express12, 2395–2414 (2022). ArticleADSMATH Google Scholar Yildirim, M. et al. Nonlinear optical feature generator for machine learning. APL Photonics8, 106104 (2023). ArticleADSMATH Google Scholar Rafayelyan, M. et al. Large-scale optical reservoir computing for spatiotemporal chaotic systems prediction. Phys. Rev. X10, 041037 (2020). MATH Google Scholar Oguz, I. et al. Programming nonlinear propagation for efficient optical learning machines. Adv. Photonics6, 016002 (2024). ArticleADSMATH Google Scholar Zhou, Y. et al. Programming the scalable optical learning operator with spatial-spectral optimization. Opt. Fiber Technol.87, 103864 (2024). ArticleMATH Google Scholar Randazzo, E. et al. Self-classifying MNIST digits. Distill5, e00027.002 (2020). Article Google Scholar Li, G. H. Y. et al. Deep learning with photonic neural cellular automata. Light Sci. Appl.13, 283 (2024). ArticleMATH Google Scholar Brown, T. B. et al. Language models are few-shot learners. Proceedings of the 34th International Conference on Neural Information Processing Systems. Vancouver, BC, Canada: ACM, 2020, 159. Hendrycks, D. et al. Measuring massive multitask language understanding. Proceedings of the 9th International Conference on Learning Representations. ICLR, 2021. Download references EPFL, Institute of Electrical and Micro Engineering, Lausanne, Switzerland Ilker Oguz, Mustafa Yildirim, Jih-Liang Hsieh, Niyazi Ulas Dinc, Christophe Moser & Demetri Psaltis You can also search for this author in PubMedGoogle Scholar You can also search for this author in PubMedGoogle Scholar You can also search for this author in PubMedGoogle Scholar You can also search for this author in PubMedGoogle Scholar You can also search for this author in PubMedGoogle Scholar You can also search for this author in PubMedGoogle Scholar Correspondence to Ilker Oguz. The authors declare no competing interests. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. Reprints and permissions Oguz, I., Yildirim, M., Hsieh, JL. et al. Resource-efficient photonic networks for next-generation AI computing. Light Sci Appl14, 34 (2025). https://doi.org/10.1038/s41377-024-01717-6 Download citation Published: DOI: https://doi.org/10.1038/s41377-024-01717-6 Anyone you share the following link with will be able to read this content: Sorry, a shareable link is not currently available for this article.
| Source: The Rosen Law Firm PAThe Rosen Law Firm PA NEW YORK, Jan. 03, 2025 (GLOBE NEWSWIRE) — WHY: Rosen Law Firm, a global investor rights law firm, reminds purchasers of common stock of MGP Ingredients, Inc. (NASDAQ: MGPI) between May 4, 2023 and October 30, 2024 (the “Class Period”), of the important February 14, 2025 lead plaintiff deadline. SO WHAT: If you purchased MGPI common stock during the Class Period you may be entitled to compensation without payment of any out of pocket fees or costs through a contingency fee arrangement. WHAT TO DO NEXT: To join the MGPI class action, go to https://rosenlegal.com/submit-form/?case_id=9167 or call Phillip Kim, Esq. toll-free at 866-767-3653 or email case@rosenlegal.com for information on the class action. A class action lawsuit has already been filed. If you wish to serve as lead plaintiff, you must move the Court no later than February 14, 2025. A lead plaintiff is a representative party acting on behalf of other class members in directing the litigation. WHY ROSEN LAW: We encourage investors to select qualified counsel with a track record of success in leadership roles. Often, firms issuing notices do not have comparable experience, resources, or any meaningful peer recognition. Many of these firms do not actually litigate securities class actions, but are merely middlemen that refer clients or partner with law firms that actually litigate the cases. Be wise in selecting counsel. The Rosen Law Firm represents investors throughout the globe, concentrating its practice in securities class actions and shareholder derivative litigation. Rosen Law Firm achieved the largest ever securities class action settlement against a Chinese Company at the time. Rosen Law Firm was Ranked No. 1 by ISS Securities Class Action Services for number of securities class action settlements in 2017. The firm has been ranked in the top 4 each year since 2013 and has recovered hundreds of millions of dollars for investors. In 2019 alone the firm secured over $438 million for investors. In 2020, founding partner Laurence Rosen was named by law360 as a Titan of Plaintiffs’ Bar. Many of the firm’s attorneys have been recognized by Lawdragon and Super Lawyers. DETAILS OF THE CASE: According to the lawsuit, defendants throughout the Class Period made materially false and/or misleading statements, and failed to disclose material adverse facts about MGPI’s business, operations, and prospects. Specifically, defendants repeatedly touted a strong demand and “normal” inventory levels in brown goods (i.e., American whiskies and tequila), when in fact there had been a slowdown in consumption and oversupply in their products. Worse, defendants had assured investors that they were positioned differently than their competitors, and that this was a non-issue, because MGPI had already taken steps to mitigate the risk, when in fact it had not. When the true details entered the market, the lawsuit claims that investors suffered damages. To join the MGPI class action, go to https://rosenlegal.com/submit-form/?case_id=9167 or call Phillip Kim, Esq. toll-free at 866-767-3653 or email case@rosenlegal.com for information on the class action. No Class Has Been Certified. Until a class is certified, you are not represented by counsel unless you retain one. You may select counsel of your choice. You may also remain an absent class member and do nothing at this point. An investor’s ability to share in any potential future recovery is not dependent upon serving as lead plaintiff. Follow us for updates on LinkedIn: https://www.linkedin.com/company/the-rosen-law-firm, on Twitter: https://twitter.com/rosen_firm or on Facebook: https://www.facebook.com/rosenlawfirm/. Attorney Advertising. Prior results do not guarantee a similar outcome. Contact Information: Laurence Rosen, Esq. Phillip Kim, Esq. The Rosen Law Firm, P.A. 275 Madison Avenue, 40th Floor New York, NY 10016 Tel: (212) 686-1060 Toll Free: (866) 767-3653 Fax: (212) 202-3827 case@rosenlegal.com www.rosenlegal.com
PLEASE NOTE: A verification email will be sent to your address before you can access your trial.
Law360 may contact you in your professional capacity with information about our other products, services and events that we believe may be of interest.
You’ll be able to update your communication preferences via the unsubscribe link provided within our communications.
We take your privacy seriously. Please see our Privacy Policy.
Law360 takes your privacy seriously. Please see our Privacy Policy.
Français 
English