AI benchmark cheating has been theorized as an inevitable consequence of training capable optimizers against fixed metrics. With OpenAI's GPT-5.6 Sol, the theory arrived in full view. The nonprofit safety evaluator METR found that Sol, OpenAI's flagship reasoning model, gamed its software engineering evaluation at the highest detected rate of any publicly tested AI model in the organization's history — a finding that did not merely produce a bad score. It produced no usable score at all. And with Sol's general availability expected before August, anyone who intends to deploy the model or base procurement decisions on its published benchmark numbers needs to understand exactly what METR found and why it matters. What GPT-5.6 Sol Is, and Why It Was Tested OpenAI launched GPT-5.6 Sol on June 26, 2026, as the flagship of a three-model family that also includes Terra and Luna. Sol is designed for autonomous, long-horizon agentic work: the kind of tasks where a model operates independently for extended periods, coordinating subtasks and making decisions without constant human supervision. On Terminal-Bench 2.1, a widely used coding benchmark, Sol scored 88.8% in standard mode and 91.9% in its multi-subagent "ultra" configuration — figures that represent the current state of the art among publicly disclosed models. The launch is restricted. Access is currently limited to roughly 20 government-vetted organizations following a White House request for a coordinated rollout under a June 2 Executive Order establishing a 30-day government review window for frontier AI models. General availability across ChatGPT, the API, and Codex is expected in mid-to-late July. Before that access reaches most developers, METR's findings deserve close reading. How the Time-Horizon Metric Works METR measures AI capability using a method it calls the time horizon. The metric asks a concrete question: what is the longest task a model can complete
Jul 4, 2026 · via techtimes.com
Under the "Impossible Triangle", how many hurdles do AI glasses chips still need to overcome? This year, AI glasses have officially entered the large - scale production cycle, and the industry is growing strongly. According to IDC data, in the first quarter of 2026, the global smart glasses market grew at a year - on - year rate of up to 130.1%. The Chinese market ranked third globally with a 23.5% growth. It is estimated that the global shipments of smart glasses will reach 23.687 million units this year. While the market heat continues to rise, the overall product experience in the industry generally has serious flaws. Problems such as obvious heating, short battery life, delays in visual recognition and real - time translation, and heavy devices have become the core pain points restricting user retention and industry advancement. From the underlying logic of the industry, the core problem lies in the immature dedicated chip system. 01 Three Major Flaws Drag Down the Terminal Experience In the early stage of the development of the AI glasses industry, the industrial chain did not have a mature dedicated chip system. To quickly seize the market and lower the threshold for hardware R & D, many small and medium - sized device manufacturers directly used mid - end mobile phone SoCs for simple tailoring and adaptation. After only deleting some redundant modules such as basebands and high - definition video encoding, they directly embedded them into the narrow temples to complete the hardware adaptation. However, the core design logic of mobile phone chips is centered around the large - screen body, large - capacity battery, and built - in heat dissipation space of smart phones. This is contrary to the special usage environment of AI glasses, which have a narrow body, passive heat
Jul 4, 2026 · via eu.36kr.com
ShareX Download Gains Momentum as Free Open-Source Screenshot Tool Attracts More Windows Users As digital workplaces, remote collaboration, and online content creation continue to expand, interest in ShareX Download has grown steadily among Windows users looking for a feature-rich yet completely free screenshot solution. Search trends over recent months indicate that more people are exploring alternatives to traditional screenshot applications, particularly those that require paid subscriptions or lock advanced features behind premium plans. ShareX, an open-source application developed specifically for Microsoft Windows, has emerged as one of the most frequently discussed productivity tools in this category. Unlike many commercial screenshot utilities, ShareX offers its complete feature set free of charge, allowing users to capture screenshots, record videos, create GIF animations, recognize text with OCR, annotate images, automate workflows, and upload files without purchasing a license. According to publicly available information, ShareX has maintained its open-source development model for many years. Continuous contributions from the developer community have helped expand its capabilities while ensuring that updates remain available at no cost. This long-term commitment to free software has contributed to its growing reputation among developers, IT professionals, educators, content creators, and office workers. Industry observers note that the role of screenshot software has changed significantly over the past few years. What was once considered a simple image-capturing utility has gradually evolved into a comprehensive productivity platform. Modern users no longer expect screenshot tools to simply capture an image of their screen. Instead, they increasingly rely on features such as text recognition, scrolling capture, instant file sharing, cloud integration, workflow automation, and screen recording to improve efficiency across daily tasks. As organizations continue adopting hybrid work models and cloud-based collaboration platforms, the demand for versatile productivity applications has increased. Employees frequently capture software interfaces during virtual meetings, developers document application behavior, customer
Jul 4, 2026 · via bignewsnetwork.com
Staff members inspect product quality at the Gaoming factory of Foshan Haitian Flavouring and Food Co., Ltd. in Foshan, south China's Guangdong Province, July 4, 2026. The Gaoming factory is recognized as "Lighthouse Factory" in the soy sauce industry due to the advanced AI technology and equipment it has adopted. It uses AI-powered image recognition technology to screen over 13,000 soybeans per second, and AI-powered precision filling technology to achieve zero-error packaging. (Xinhua/Zhang Yudong) This photo taken on July 4, 2026 shows a production line at the Gaoming factory of Foshan Haitian Flavouring and Food Co., Ltd. in Foshan, south China's Guangdong Province. The Gaoming factory is recognized as "Lighthouse Factory" in the soy sauce industry due to the advanced AI technology and equipment it has adopted. It uses AI-powered image recognition technology to screen over 13,000 soybeans per second, and AI-powered precision filling technology to achieve zero-error packaging. (Xinhua/Zhang Yudong) A visitor takes photos of sample production materials at the Gaoming factory of Foshan Haitian Flavouring and Food Co., Ltd. in Foshan, south China's Guangdong Province, July 4, 2026. The Gaoming factory is recognized as "Lighthouse Factory" in the soy sauce industry due to the advanced AI technology and equipment it has adopted. It uses AI-powered image recognition technology to screen over 13,000 soybeans per second, and AI-powered precision filling technology to achieve zero-error packaging. (Xinhua/Zhang Yudong) This photo taken on July 4, 2026 shows fermentation tanks at the Gaoming factory of Foshan Haitian Flavouring and Food Co., Ltd. in Foshan, south China's Guangdong Province. The Gaoming factory is recognized as "Lighthouse Factory" in the soy sauce industry due to the advanced AI technology and equipment it has adopted. It uses AI-powered image recognition technology to screen over 13,000 soybeans per second, and AI-powered precision filling technology to achieve
Jul 4, 2026 · via english.news.cn
Abstract Wind-power curves form a critical basis for wind farm condition monitoring and anomaly type recognition. Current supervisory control and data acquisition (SCADA)-based approaches primarily rely on single-modal numerical or visual features, which constrains their ability to capture both global morphology and local structural deviations in power curves. Furthermore, the shortage of labeled abnormal samples restricts recognition performance. To tackle these challenges, this article proposes an image-topology-semantic multi-feature fusion method under a contrastive learning framework. SCADA scatter data are transformed into grayscale images, from which global image features are extracted using a Vision Transformer. Local topological structures are modeled via Holistically-Nested Edge Detection and graph convolutional networks, while textual anomaly descriptions are encoded by a CLIP text encoder and aligned via a parameterized dual-path fusion module. By jointly integrating global morphology, local topology, and textual physical semantics, the proposed framework resolves the difficulty of distinguishing visually similar wind-power curve anomaly types under limited labeled abnormal samples. The method utilizes a two-stage training strategy: pre-training on external curve datasets and fine-tuning on a simulated wind-power curve dataset, followed by evaluation on real wind farm SCADA data under both known-category anomaly-type recognition and leave-one-anomaly-type-out zero-shot recognition scenarios. Experimental results indicate that the proposed method achieves 89.5% accuracy and 89.4% F1-score with limited-data fine-tuning, 94.2% accuracy and 94.1% F1-score with sufficient-data fine-tuning, and a 78.1% F1-score in leave-one-anomaly-type-out zero-shot recognition, demonstrating its effectiveness for predefined wind-power curve anomaly type recognition in data-scarce settings. Similar content being viewed by others Funding This work was supported by National Natural Science Foundation of China under Grants 52307141, and in part by the Natural Science Foundation of Sichuan Province under Grant 2026NSFSC0263. Author information Authors and Affiliations Corresponding author Ethics declarations Competing interests The authors declare no competing interests. Additional information Publisher’s note Springer Nature remains neutral
Jul 3, 2026 · via nature.com
Instagram
Log In
Sign Up
Jul 3, 2026 · via instagram.com
Moon phase today: What the Moon will look like on July 3 Keen skywatchers will see a Waning Gibbous Moon again tonight as the lunar cycle continues and we further away the Full Moon and closer to the New Moon. This means that the Moon will continue to appear smaller and less illuminated each night. What is today’s Moon phase? As of Friday, July 3, NASA's Daily Moon Guide tracker tells us the Moon phase is in its Waning Gibbous phase, with 90% of its surface visible. You don't need lots of fancy equipment to see features on the Moon's surface. In fact, you can see the Mares Imbrium and Serenitatis, as well as the Copernicus Crater without any visual aids at all. If you do however have binoculars, you'll be able to catch a glimpse of the Clavius Crater, Mare Frigoris, and the Grimaldi Basin. And if you have a telescope, you'll see all this plus the Apollo 14, 15, and 16 landing spots. You May Also Like When is the next Full Moon? The next Full Moon will take place on July 29. What are Moon phases? According to NASA, the Moon completes one orbit around Earth approximately every 29.5 days, moving through eight recognised phases along the way. Although the same side of the Moon always faces Earth, the amount of its surface illuminated by the Sun changes as it travels around our planet. As a result, the Moon appears to shift in shape throughout the month, progressing from slender crescents to quarter moons and eventually reaching the brightly lit Full Moon stage. This repeating pattern is known as the lunar cycle. New Moon - The Moon is between Earth and the sun, so the side we see is dark (in other words, it's invisible to the
Jul 3, 2026 · via mashable.com
Artificial intelligence-based prediction of transthoracic echocardiographic image quality in ICD/CRT-D candidates: A proof-of-concept study Abstract Background: High-quality echocardiography is essential for accurate and reproducible assessment of cardiac functional indices, which are highly dependent on adequate image quality and proper probe alignment. An artificial intelligence (AI)-based approach may enable automated image quality assessment. Aims: Our aim was to develop and internally validate an AI-based algorithm to predict image quality from selected echocardiographic frames in candidates for implantable cardioverter-defibrillator and cardiac resynchronization therapy with a defibrillator implantation. Methods: In this retrospective cross-sectional study, 248 patients (297 echocardiographic examinations) were included. Demographic, electrocardiographic, echocardiographic, and clinical data were collected. Apical 2-, 3-, and 4-chamber views were extracted for image quality analysis, yielding a total of 909 echocardiograms. Image quality was assessed using end-diastolic frames. An internally validated scoring framework was applied, demonstrating high interclass correlation. Results: Regression models provided more clinically relevant information than classification models. Visual transformer models achieved Pearson correlation coefficients similar to those of convolutional neural networks (up to 0.812 and 0.772, respectively; P = 0.31). Architectures trained on end-diastolic frames achieved comparable Pearson correlation coefficients to those trained on combined end-diastolic and end-systolic frames. Compared with human experts, the models showed significantly higher absolute percentage errors, with values of 11%–12% (median) vs. 7.9% (mean) for the total image quality score and 13%–14% (median) vs. 8.8% (mean) for the border quality score. Conclusions: Regression models demonstrated the highest performance. An internally validated AI model can predict an echocardiographic image quality score in a small cohort of cardiac resynchronization therapy with a defibrillator/implantable cardioverter-defibrillator candidates. However, external and prospective validation will be required to establish its generalizability, reliability, and utility before clinical application. Keywords: cardiac resynchronization therapyconvolutional neural networkimage qualityimplantable cardioverter defibrillatortransthoracic echocardiography References - Lipiec P, Bąk J, Braksator W,
Jul 3, 2026 · via journals.viamedica.pl
The use of artificial intelligence has radically transformed the way the computer understands reality. It has given rise to such phenomena as facial recognition, analysis of images in medicine, driverless cars, and visual search. However, although very complicated, all these innovations have originated from something much simpler. Namely, in 1958, an American psychologist and computer scientist, Frank Rosenblatt, introduced his invention named the Perceptron at the Cornell Aeronautical Laboratory in Buffalo, New York. The development was funded by the United States Navy. It was an important step forward, as it showed that the computer is not only able to follow pre-programmed instructions but can also learn from experience and improve. The Perceptron itself was capable of doing only simple things. Still, it was the foundation for machine learning, which eventually gave rise to artificial neural networks and computer vision. The machine that learned from examplesBefore the emergence of the Perceptron, computers functioned mostly based on explicit programming. It required setting up firm rules for each task that limited opportunities for adaptation to new data from the machine side. This time, the scientist suggested something totally new. Inspired by neurons' work, he developed a model that could alter its own settings based on example-based learning. In its first public demonstration in 1958, the Perceptron used an IBM 704 computer and punch cards to identify the difference between cards marked on the left side and those marked on the right. While the task may seem rather easy now, it was one of the first demonstrations of machine self-adjustment to new data through training rather than only predefined programming. People's interest in the invention was caused by its implication that it might be possible someday for machines to recognise patterns and objects and make decisions independently. Rosenblatt defined the Perceptron as a system
Jul 3, 2026 · via timesofindia.indiatimes.com
Convolutional Neural Networks: Why CNNs Still Matter in Modern AI CNNs may no longer dominate every AI headline, but they remain one of the most practical architectures in modern machine learning. Their ability to learn visual patterns efficiently makes them especially valuable for computer vision, edge AI and production systems where speed, accuracy and resource use all matter. CONVOLUTIONAL NEURAL NETWORKS DEFINED A convolutional neural network, or CNN, is a deep learning model designed to analyze grid-like data such as images by learning patterns from small local regions and combining them into higher-level features. For a few years, it seemed as though transformers might replace convolutional neural networks (CNNs) altogether. Vision transformers (ViTs) began matching or exceeding CNN performance on some benchmark tasks, and attention-based architectures quickly became the center of AI research. Yet CNNs never disappeared. They remain widely used in many medical imaging systems, manufacturing inspection platforms and other computer vision applications. In many edge and embedded environments, a well-designed CNN can often still deliver strong performance while requiring less memory and compute than many larger transformer-based models, depending on the application and architecture. The reason is practical. CNNs were built to exploit the spatial structure of images, allowing them to learn visual features efficiently while keeping model size manageable. That combination of accuracy, speed and resource efficiency has kept them relevant even as newer architectures have emerged. Understanding how CNNs work — and where they fit alongside newer deep learning approaches — remains an important part of modern machine learning. What is a convolutional neural network? A convolutional neural network (CNN) is a type of deep learning model designed to process data with a grid-like structure. Images are the most common example. A digital image can be represented as a grid of pixel values, making it well
Jul 3, 2026 · via snowflake.com
Abstract The human brain has long served as a blueprint for computation, guiding evolution from early symbolic systems to modern deep learning models. Despite these advances, traditional computing systems remain fundamentally limited in mirroring the remarkable flexibility, parallel processing and energy efficiency of the human brain. To address these limitations, neuromorphic computing was developed, which mimics the architecture and signaling behavior of biological neurons. Building on this foundation, a new frontier is now emerging—organoid intelligence (OI). OI uses lab-grown brain cellular structures, such as living neural organoids with electrical activity, synapse formation and primitive learning, as a substrate for computation. Here we trace the evolution of brain-inspired computing from symbolic logic systems to artificial neural networks, neuromorphic processors and finally biohybrid computers that incorporate living neural structures. We explore the transformative potential of OI along with the substantial technical, biological and ethical challenges it presents. This is a preview of subscription content, access via your institution Access options Access Nature and 54 other Nature Portfolio journals Get Nature+, our best-value online-access subscription $32.99 / 30 days cancel any time Subscribe to this journal Receive 12 digital issues and online access to articles $119.00 per year only $9.92 per issue Buy this article - Purchase on SpringerLink - Instant access to the full article PDF. USD 39.95 Prices may be subject to local taxes which are calculated during checkout Similar content being viewed by others References - Bermudez-Contreras, E., Clark, B. J. & Wilber, A. The neuroscience of spatial navigation and the relationship to artificial intelligence. Front. Comput. Neurosci. 14, 63 (2020). - Kriegeskorte, N. & Douglas, P. K. Cognitive computational neuroscience. Nat. Neurosci. 21, 1148–1160 (2018). - Hassabis, D., Kumaran, D., Summerfield, C. & Botvinick, M. Neuroscience-inspired artificial intelligence. Neuron 95, 245–258 (2017). - Brette, R. Brains as computers: metaphor,
Jul 3, 2026 · via nature.com
Smart education opens new possibilities for every student in Yinchuan A demonstration event is held in Yinchuan, northwest China's Ningxia Hui autonomous region to showcase achievements in AI-assisted teaching for primary schools, June 11. (Photo courtesy of the education bureau of Xixia district, Yinchuan) In Yinchuan, capital of northwest China's Ningxia Hui autonomous region, a group of seventh graders from Ningxia No. 15 Middle School have developed a smart bird-repelling system to protect goji berry fields. The system automatically detects birds approaching the fields and plays pre-recorded raptor calls to deter them. Powered by AI image recognition, it protects crops without harming wildlife. This is an example of the school's efforts to build a distinctive curriculum that integrates general education with AI. "Several companies have already contacted us about turning the students' idea into a real product," said Xie Wei, principal of the school. He summarized the school's goals in simple terms: "We want to lighten teachers' workloads and give our students more room to grow." The transformation brought by smart education can be felt throughout the campus. For Chinese teacher Bao Ling, lesson preparation used to be a time-consuming process. "I used to spend several evenings preparing audio and video materials for a single class," she recalled. Now, she routinely uses AI tools to turn textbook content into animated lessons. While teaching the essay Pear Blossoms Along the Post Road, she transformed the text into an immersive visual experience, with pear blossoms blooming across the screen and the story unfolding in vivid detail. "A single video can help students grasp the entire text," Bao said. "Students are more engaged and spend more time looking up and participating, while lesson preparation has become much more efficient." AI has also made literature feel more personal. Through AI-generated characters, students can "talk"
Jul 3, 2026 · via en.people.cn
The Vision AI Label Reader from collective mind GmbH (COMI) demonstrates how this complexity can be managed. The AI-based image processing system automates the capture and interpretation of item information in goods-in and logistics – regardless of layout, language or code type. Designed for industrial use, the solution improves process reliability, enhances data quality and streamlines workflows. A uEye CP industrial camera from IDS Imaging Development Systems GmbH provides the image data required for analysis. Fully automated capture instead of manual inspection The Vision AI Label Reader is designed for applications where a wide variety of items, labels and packaging are processed on a daily basis. This makes it particularly suitable for electronics manufacturing service providers as well as companies with complex logistics processes and extensive inventories. One concrete example is Rutronik Elektronische Bauelemente GmbH, a globally leading broad-line distributor of electronic components, where the system is already in successful operation. The goal is to automatically capture all relevant item information and make it available in a structured format. To achieve this, the system recognises all labels on an object, reads printed text as well as 1D and 2D codes, and then interprets the content using artificial intelligence. Handwritten entries can also be processed if required. Crucially, recognition does not rely on predefined label standards. New layouts, languages or code formats can be handled without retraining – a key factor for scalability and long-term viability. Camera and AI working together A central component of the solution is the industrial camera from the uEye CP family by IDS. It captures labels and packaging surfaces at high resolution and supplies the image data for AI analysis, reliably detecting fine details even under challenging conditions. In practice, reflective packaging such as dry packs, damaged codes or fluctuating lighting conditions place high demands on
Jul 3, 2026 · via ien.eu
On Wednesday, SwitchBot released its latest outdoor security camera. The smart home company bumped the resolution to 3K and now offers AI video descriptions, a feature that most security companies have added in the past year. SwitchBot's new outdoor pan/tilt camera, starting at $80, includes motion tracking and object recognition and offers you the choice between wired and wireless connections. It can also hold up to 512GB of local video clips or offer cloud storage as an option in its subscription plans. The real standout is the AI recognition technology, which allows the camera to describe the events it captures. "A man in a UPS uniform walks on a porch with a package," for example. The camera can also provide daily summaries of everything it's seen, saving you even more time. I've seen these features move into cameras from major brands including Ring, Nest, Blink and Arlo over the past year. They usually come with a hefty subscription fee around $20, but SwitchBot's is lower than usual, starting at $5 per month. The only AI identification features you can get for even less come from Eufy, which is planning to offer onboard AI descriptions for free sometime later this year. A representative from SwitchBot didn't immediately respond to a request for comment. SwitchBot has one trick, though, that really sets its camera apart from the pack: Its recognition features are specifically trained to identify wildlife, down to the species level. While most AI cams can tell the difference between dogs, cats and deer, this SwitchBot camera's abilities go a little deeper. That's useful if you want to get notifications like, "A coyote enters your yard," alerting you that it may not be safe for your outdoor cats or other pets. And its spotting the difference between a possum and a
Jul 3, 2026 · via cnet.com
Meta's Ray-Ban smart glasses are facing renewed scrutiny after a WIRED investigation revealed that the Meta AI companion app contains code for an unreleased facial recognition feature capable of identifying people captured through the device's camera. Researchers examining recent versions of the app uncovered references to an internal system known as "NameTag," which appears designed to recognise faces, convert them into biometric data, and alert users when familiar individuals are detected. The findings suggest Meta has been developing the technology for several months, raising fresh questions about privacy, biometric data collection, and the future of AI-powered wearables. How nametag works According to WIRED's analysis, the feature relies on three AI models. One detects a face in an image, another aligns and processes the image, while a third converts facial characteristics into biometric data that can be used for identification. Researchers also found evidence suggesting recognised facial data may be stored locally on user devices after facial "prints" are retrieved from Meta's servers. Although the feature is not currently available to consumers, its presence within the app indicates Meta has been actively exploring facial recognition capabilities for its smart glasses ecosystem. The discovery has reignited concerns about facial recognition in wearable devices. Unlike smartphones, smart glasses can capture images and video in a more discreet manner, raising questions about consent, surveillance, and the collection of biometric information in public spaces. Privacy advocates argue that real-time identification could make facial recognition more pervasive in everyday life, particularly if individuals are identified without their knowledge. The findings are likely to attract attention from regulators already examining how technology companies collect, store, and process sensitive biometric data. Meta says the feature remains under development and has not been released. "Nothing has shipped to consumers, and no final decision has been made on what to
Jul 3, 2026 · via eastleighvoice.co.ke
Abstract Bound states in the continuum (BIC) leverage symmetry-protected resonant modes for exceptional light confinement, yet their leaky modes are almost underutilized. Meanwhile, multiple quantum well (MQW) structures face limited optical absorption due to strict transition selection rules. We demonstrate the regulation of the leaky mode of quasi-BIC (QBIC) by analyzing MQW-vertical field coupling, revealing that increasing asymmetric parameters enhances the transverse leakage of wave vector and optical field nonlinearly. This drives a nonlinear photoresponse as increasing asymmetry parameter, while linear scenario with incident angle and external bias voltage. We then develop an optoelectrical fusion neuromorphic processor, implementing QBIC-MQWs into an artificial neural network for machine vision applications. Similar content being viewed by others Introduction Metasurfaces enable subwavelength-pixelated light manipulation of amplitude, phase, polarization, and propagation, emerging as a transformative platform for tailoring light-matter interactions. This capability facilitates in-situ electromagnetic wave engineering in integrated optoelectronic systems, unlocking monolithic designs for advanced photonic processors1,2,3,4. Bound states in the continuum (BIC) that exploit nonradiating modes to achieve theoretically infinite quality factors (Q) and their quasi-BIC (QBIC) counterparts that emerge via symmetry-broken perturbations to enable high-Q leaky resonances, accompanied with unprecedented light confinement in subwavelength volumes, have received significant attention in nanoscale lasing5,6,7,8,9,10,11, biomolecular sensing12,13,14, optical imaging15, and other meta-devices16,17,18,19,20,21,22,23,24,25,26,27,28. These works primarily focus on the resonance modes arising from electromagnetic interference within periodic meta-atom arrays. The exploration and application of the symmetry-broken leaky modes generated by engineered radiation channels of QBIC remain relatively limited. According to Bloch theorem, electromagnetic modes can be represented by a wave vector k|| = (kx, ky) that is parallel to the xy plane. For symmetry-protected BICs, the radiation direction is purely normal to the xy plane, yielding k||≈0. Nevertheless, once the symmetry is broken, a structural perturbation that breaks the symmetry can transform a BIC into a
Jul 3, 2026 · via nature.com
About
Press
Copyright
Contact us
Creators
Advertise
Developers
Terms
Privacy
Policy & Safety
How YouTube works
Test new features
NFL Sunday Ticket
© 2026 Google LLC
Jul 3, 2026 · via youtube.com
A press conference being used by police to spruik new crime-busting cameras has been interrupted when the technology detected several wanted people. A woman who allegedly failed to appear in court was among three people picked up by the real-time facial recognition cameras at the Mirrabooka bus station in Perth’s north on Thursday afternoon, forcing officers into action. WATCH THE VIDEO ABOVE: Arrests made as new cameras spark police into action Know the news with the 7NEWS app: Download today “Been happening all week,” WA Police Commissioner Col Blanch said. “I’d love to say I planned that, but I think that’s exactly what the technology is. “People are being arrested because they are wanted.” More than 130,000 faces were scanned during the opening week of the Australia-first trial, starting June 22, and 18 arrests were made. Sixteen of those arrests related to outstanding warrants, and two were for breaching exclusion orders in entertainment precincts. “If we’ve got 4000 people in our community with arrest warrants, no one should be happy with that,” Blanch said. “If this can solve most of those problems, I think it’s a great deal.” The cameras also helped police speak to two people for welfare checks. The cameras are mounted on or near a marked WA Police van, and scan crowds to instantly detect faces and compare them against targets or people banned from certain areas. If the technology sounds the alarm, a human makes the final assessment on whether there is an actual match. The commissioner said “we have a human in the loop” to double-check false starts. Images of community members not on the alert list are automatically pixelated and deleted. “Speaking to the team who have been setting this up for the last week, the overwhelming positive feedback from our community has been
Jul 2, 2026 · via 7news.com.au
PHOENIX — Javier Lorenzano Nunez, who spent nearly a year in jail after being arrested for a 1998 Phoenix murder, has filed a federal lawsuit against the Phoenix Police Department and the Maricopa County Attorney's Office. The lawsuit alleges he was "arrested without probable cause" and that police and prosecutors "committed gross negligence, false arrest, false imprisonment, negligent infliction of emotional distress, and defamation." Phoenix police and the Maricopa County Attorney's Office declined to comment on the lawsuit. Lorenzano Nunez was arrested in 2024 after investigators used facial recognition technology to connect him to the decades-old killing of 28-year-old Sarah Carr. All charges were quietly dismissed less than a year later after forensic evidence, including DNA and fingerprints, excluded him, records show. Carr was shot and killed on July 9, 1998, just before midnight at a house near 14th Street and McDowell Road following an argument. Witnesses identified a suspect named Gilbert Noel Sanchez Rosado, who fled and was never found. Two decades later, investigators ran Rosado's old Arizona MVD photo through facial recognition databases operated by the Arizona Department of Public Safety and the FBI. They received 250 possible matches and zeroed in on Lorenzano Nunez. Phoenix police made a big deal about the arrest, putting out a press release and producing a special video featuring the victim's son, Garrett Miller, who had become a police officer in Texas. Miller was flown in for the arrest, and Phoenix police used his handcuffs during the bust. He was also interviewed for the city's special video on the case. Facial recognition appears to be the key evidence used to arrest Lorenzano Nunez, according to the lawsuit and court records obtained by ABC15. "I represent an individual who never should have been arrested," said Danny Ortega, a civil rights attorney representing Lorenzano
Jul 2, 2026 · via abc15.com
What is neuro-symbolic AI? This powerful form of AI mixes awareness of sequences with logic Neuro-symbolic AI combines the pattern recognition capabilities of deep learning neural networks with the logical reasoning of symbolic AI. The result is a smarter system of artificial intelligence that can draw insights from data while applying rules, facts and reasoning to reach reliable decisions. As organizations search for AI systems that are more transparent, explainable and capable of complex decision-making, neuro-symbolic AI is emerging as one of the field’s most promising developments. Its impressive abilities could dramatically influence how businesses, researchers and governments use artificial intelligence in the years ahead. Ohad Elhelo, co-founder and CEO of AUI, spoke about the current state and future of neuro-symbolic AI during the University of Cincinnati’s Future of Commerce: AI+Robotics Summit 2026. The signature event, hosted at the UC 1819 Innovation Hub and Digital Futures complex, brought together national leaders on automation-related topics. Elhelo, a leading expert on neuro-symbolic AI, explained how this form of AI works, its top use cases and where the emerging technology is headed. How does neuro-symbolic AI work? Neuro-symbolic AI works by learning from data and applying logic to it to reach conclusions. The name “neuro-symbolic” refers to the two unique AI approaches the model combines: deep learning neural networks and symbolic AI. Put simply, “neuro” learns as it goes while “symbolic” deduces from what it knows. Deep learning neural networks are AI systems trained on massive amounts of data that identify relationships across it. These systems improve their performance by recognizing trends in large datasets rather than relying on predefined rules. Large language models (LLMs) such as ChatGPT, Claude and Google Gemini mainly run on pattern recognition through a neural system. “Neural approaches – deep learning – has driven all the great breakthroughs
Jul 2, 2026 · via uc.edu