We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.If youve been eyeing an iPad for work or creative projects, this 13-inch iPad Air (wifi + cellular) is currently on sale for $999.97, down from $1,249the lowest price its ever been, according to price trackers. At 11.04 by 8.46 by 0.24 inches and 1.36 pounds, this is reportedly the heaviest iPad yet, but it still feels portable. Additionally, its aluminum chassis gives it a more premium, sturdy feelthough it lacks an IP rating, so maybe dont put it to the test near water. For connectivity, the cellular model skips the physical SIM card tray in favor of eSIM-only support.This iPad impresses with its 2,732 x 2,048 resolution, 600 nits of peak brightness, True Tone for white balance, and LED-backlit IPS LCD with P3 color support. However, creators who value OLEDs deeper blacks and higher contrast might find the LCD screen subpar. That said, the Airs display handles glare well and resists fingerprints effectively. Audio is another highlight, with its stereo speakers delivering rich bass, especially in landscape mode, enhancing the experience for movies and music. The front cameras new placement along the long edge is a welcome change, making video calls feel more natural. Speaking of, both the front and rear 12MP cameras of the Air produce sharp, vibrant images, though performance in low-light settings could be better.Powered by the M2 chip and running on iPadOS, the Air feels snappy (even with its 60Hz refresh rate)multitasking is smooth, and tasks like editing in iMovie or using GarageBand are nearly as fast as on the Pro, notes this PCMag review. And, with 512GB of storage and 8GB RAM, its more than capable of handling app switching, media streaming, and general use with ease. Battery life is standard Apple fare (about 7 hours in real-world use) and charging caps at 20W. As for accessories, the Air is compatible with the Magic Keyboard ($319, down from $349) and Apple Pencil Pro ($114.99, down from $129), adding flexibility for work and/or creativity.

From Ethernet to USB-C power: Lenovos AI-assisted storage device showcases design efficiency and hints at a game-changing ThinkStorage evolution.

The laptop mainboard will be made by RISC-V developer DeepComputing.

There are times you dont need an intricately designed light source to illuminate your surroundings or to add a certain flair to your space. Sometimes you just need a well-thought of kind of lamp that is simple and minimalistic but can also meet your functional needs. Designers take their inspiration from a variety of things but this new and unconventional lamp is inspired by a mundane thing: the store sign.Designer: Amber DewaeleThe Flat Light is exactly what its name says. It is housed in a thin, flat rectangular form that uses stainless steel and acrylic materials. The light flows through the edges so you can expect just a subtle glow to add to the atmosphere or to the aesthetic of your space. Since it has a compact shape, it can fit into narrow spaces and can also become a wall-mounted light.The light comes in two kinds of finishes. The brushed finish has a more fine texture and gives you a semi-matte surface. The polished finish meanwhile has a double purpose since it can also act as a mirror to reflect on your surroundings. It is not wireless though so you still have to place it somewhere that has a nearby power source.Its an interesting kind of lamp that acts more like part of your furniture or surroundings rather than a full on light source. It will fit perfectly on your dresser, make-up space, bookshelf, picture shelf, and other places where you may need some small source of light, with a minimalist but fashionable twist.The post Flat Lamp gives you a minimalist and aesthetic light source first appeared on Yanko Design.

Ubisofts open-world action RPG Assassins Creed Shadows has been delayed again, launching on March 20th instead of February 14th. While the developer specified it requires a few more weeks for additional polish, the rumored preview events are still happening.Insider Gamings Tom Henderson reaffirmed this in a recent tweet and later reported that the embargo is January 23rd, 6 PM CET. However, media outlets and content creators can seemingly reveal their participation on January 16th. These previews were reportedly meant for late September last year but were cancelled before the games first delay announcement.Launching for Xbox Series X/S, PS5, and PC, Assassins Creed Shadows has received blog posts and gameplay snippets outlining stealth, combat and parkour. Details on the first expansion, Claws of Awaji, have also leaked courtesy of its Steam page. It introduces a new region, enemy types, skills, and gear, offering over ten hours of content. Head here for more details.Despite the delay of Assassins Creed Shadows, the preview events are still going ahead for media and creators next week.Its unclear when previews will be lifted, but I will know soon! pic.twitter.com/n8nZYqY3ti Tom Henderson (@_Tom_Henderson_) January 10, 2025The embargo is January 23rd, 6PM CET. Media can announce they played the preview on the 16th Tom Henderson (@_Tom_Henderson_) January 10, 2025

While Assassins Creed Shadows has suffered another delay, plans are still proceeding apace. A new batch of previews is reportedly going live on January 23rd. However, details of the first expansion have been revealed before that courtesy of the games Steam page.According to the now-pulled description, its titled Claws of Awaji and features a new region to explore (one where players will feel the fear of being hunted and under threat). It also introduces a new weapon type in the Bo, new skills and gear, and fresh enemies to deal with as players search for lost treasure. Perhaps the most notable part of the expansion is its promise of over ten hours of new content.Claws of Awaji will launch later this year for free to those who pre-order the game. Ubisoft may offer more details shortly before launch, so stay tuned for updates.Assassins Creed Shadows launches on March 20th for Xbox Series X/S, PS5, and PC. Its recent delay was due to the developer requiring a few additional weeks to implement feedback and ensure an even more ambitious and engaging day-one experience.

Rebellion Developments next title, Sniper Elite: Resistance, is out January 30th, but its next big release, Atomfall, is arguably more intriguing. The post-nuclear survival action title has received a new overview trailer, setting the stage for the story, protagonist, and what they can expect. Check it out below.Occurring years after the Windscale fire in 1957, the player awakens with no memories near a quarantined village. As they deal with its overbearing occupying force and B.A.R.D robots, they must interact with NPCs, follow leads (in the absence of a quest system), and unearth clues to their past. Of course, the world is far from peaceful between the cults, B.A.R.D. Bunkers and mutants.Players can discover and upgrade different weapons, from cricket bats to rifles. Ammo can be hard to come by, which sometimes means crafting your own. Other times, you must scrounge for supplies (which is when the metal detector comes in handy).Atomfall launches on March 27th for Xbox One, PS4, PS5, Xbox Series X/S, and PC alongside Game Pass. Deluxe Edition owners get three days of advanced access head here for more details.

As one of 6,000 journalists at the show, I walked around a lot to find the coolest tech. My conclusion: The robots are coming.Read More

At this point, its pretty clear what Donald Trump wants from Mark Zuckerberg. But what does Zuckerberg, who has now gone to Mar-a-Lago twice since the November election, want from the President-elect?Thats the question Ive been asking sources in and around Meta over the last several days. They all described Metas relationship with the outgoing Biden administration as incredibly hostile. Its safe to assume that Zuckerberg wants a reset for the MAGA regime, especially since Trump threatened not that long ago to imprison him for life.In Trumps America, removing tampons from the mens restrooms on Metas campuses, a real thing that just happened is as much a business decision as a political one. Destroying woke ideology is a key pillar of Trumps stated mandate. Others who know they need to play the game, like Amazon, are also starting to fall in line. Even still, Zuckerberg is transforming Meta for this new political reality at a speed thats unusual for a company of its size and influence. Founder mode.In his conversation with Joe Rogan and his video on Instagram, Zuckerberg shares a laundry list of issues that Trump could help him with: fighting other countries that are ratcheting up their policing of his platforms, stopping Apple from dictating how he builds mobile apps and smart glasses (the latter is increasingly important to Metas future), and, perhaps most importantly, keeping domestic AI regulation from slowing his efforts to crush OpenAI. Elon Musk has bought Trumps ear. But the more time Zuckerberg spends in Mar-a-Lago, the more Sam Altman and Tim Cook should be worried.Then theres the US governments case to break up Meta thats set to go to trial in a few months. After the blur that was the last four years, its easy to forget that this lawsuit was filed at the end of Trumps first term by a Republican FTC chair, not Lina KhanMost of the headline reactions from the past week have focused on Zuckerbergs decision to end Metas third-party fact check program. It was a convenient scapegoat for company executives that, frankly, never lived up to the goal of bringing more neutrality to Facebook and Instagram. The Community Notes alternative Meta is cribbing from X was not on the product roadmap before this week, so it will probably be awhile before everyone sees it in the wild.The announcement that US moderators would be moved from California to Texas is perhaps the most cynical of them all; talk to anyone who knows and theyll tell you the vast majority of moderators are already based in Austin.Do you work at Meta? Id love to chat. You can reach me me securely on Signal and Ill keep you anonymous.The hateful speech that is now allowed on Metas is eye-popping on its face and will be deserving of more scrutiny in the coming weeks. The decision to start recommending political content again is a 180-degree turn for Zuckerberg. But insiders believe that the most impactful change for users of Metas apps will be the softening of its systems that remove content for potential policy violations.Out of all the announcements Meta made last week, this is the one I believe is the least connected to Trump. Meta execs have been signaling for a while that they know they are mistakenly removing too much content that doesnt actually break the rules; Im told its one of, if not the, biggest complaint in user surveys. If done correctly, dialing back on moderation mistakes may be the only thing Zuckerberg announced that makes everyone happy.ElsewhereCES is for dealmaking now: Each year, the official CES show the sprawling show floor and flashy keynotes feels more like an advertising exercise and no longer a place to launch real products. Most of the energy has moved to private meeting rooms and happy hours at the Wynn, Aria, and Cosmopolitan, where tech execs are schmoozing CMOs and getting deals done with partners all week. At this shadow CES, everyone seems to agree that the show is more alive than ever. Booths on the show floor have become marketing tools to show clients before you take them to a steak dinner. The challenge for the organizers of CES will be figuring out how to bridge the growing influence of this part of the show with their current business model of charging people to walk around booths filled with smart toasters and concept cars.TikTok may just get banned: Imagine an alternate world in which the Chinese government is about to ban Instagram from operating in the country and Mark Zuckerberg is in hiding. Thats the situation with ByteDance and its founder Zhang Yiming, who stepped down from the CEO role after the last US ban attempt but still controls the company. He let TikTok be banned in India and seemingly has no interest in the app surviving this time, so why wouldnt he let the same thing happen again?Google and OpenAI flick at whats next: Googles DeepMind unit is starting an ambitious project to build generative models that simulate the physical world, which it believes is on the critical path to artificial general intelligence. Meanwhile, OpenAI is returning to its early roots by starting a general-purpose robotics team that will build hardware and push towards AGI-level intelligence in dynamic, real-world settings. We may have hit a scaling wall on text data but the big labs clearly see an opportunity in 3D. (See also what Nvidia announced last week.)Other headlines you may have missed: Tencent (a large investor in Epic Games, Snap, and US tech companies)put on the Pentagons blacklist for being allegedly under the influence of the Chinese military. Tim Cooks totalrose 18 percent last year to $74.6 million. Elon Musk is hosting an inauguration party for Trump in DC with Uber and The Free Press. Sam Altmans sister filed a sexual abuse lawsuit against him.Job boardSome recent, noteworthy job changes in the tech world:A bunch of changes at Meta: UFC CEO Dana White, Exor CEO John Elkann, and Charlie Songhurst joined the board. Joel Kaplan is running policy and comms now. After a stint at Google, Im told Michael Levinson is coming back as VP of product for the Integrity org. (Good luck!) Head of civil rights, Roy Austin, is leaving. And former DEI chief Maxine Williams is now head of accessibility and engagement.Elon Musks X named a couple of new leaders: Romina Khananisho is the new head of government affairs and John Nitti is head of ad innovation.Calista Redmon joined Nvidia as VP of global AI initiatives, where shell drive adoption of the NVIDIA platform for national and regional AI initiatives.Sophia Dominguez, Snaps director of AR platform, is leaving.More linksMy colleagues rounded up the best things they saw at CES this year.Kylie Robison on how Elon Musks xAI is quietly taking over X.Ethan Mollick sounds the alarm about what the AI labs are claiming is coming.Eric Goldmans 2024 Internet Law Year-in-Review.Ming-Chi Kuo is bearish on Apple this year.Derek Thompson on the anti-social century.If you havent already, dont forget to subscribe to The Verge, which includes unlimited access to Command Line, all of our reporting, and an improved ad experience on the web.As always, I want to hear from you, especially if you work at Meta. Respond here, and Ill get back to you, or ping me securely on Signal.Thanks for subscribing.

Autoregressive pre-training has proved to be revolutionary in machine learning, especially concerning sequential data processing. Predictive modeling of the following sequence elements has been highly effective in natural language processing and, increasingly, has been explored within computer vision domains. Video modeling is one area that has hardly been explored, giving opportunities for extending into action recognition, object tracking, and robotics applications. These developments are due to growing datasets and innovation in transformer architectures that treat visual inputs as structured tokens suitable for autoregressive training.Modeling videos has unique challenges due to their temporal dynamics and redundancy. Unlike text with a clear sequence, video frames usually contain redundant information, making it difficult to tokenize and learn proper representations. Proper video modeling should be able to overcome this redundancy while capturing spatiotemporal relationships in frames. Most frameworks have focused on image-based representations, leaving the optimization of video architectures open. The task requires new methods to balance efficiency and performance, particularly when video forecasting and robotic manipulation are at play.Visual representation learning via convolutional networks and masked autoencoders has been effective for image tasks. Such approaches typically fail regarding video applications as they cannot entirely express temporal dependencies. Tokenization methods such as dVAE and VQGAN normally convert visual information into tokens. These have shown effectiveness, but scaling such an approach becomes challenging in scenarios with mixed datasets involving images and videos. Patch-based tokenization does not generalize to cater to various tasks efficiently in a video.A research team from Meta FAIR and UC Berkeley has introduced the Toto family of autoregressive video models. Their novelty is to help address the limitations of traditional methods, treating videos as sequences of discrete visual tokens and applying causal transformer architectures to predict subsequent tokens. The researchers developed models that could easily combine image and video training by training on a unified dataset that includes more than one trillion tokens from images and videos. The unified approach enabled the team to take advantage of the strengths of autoregressive pretraining in both domains.The Toto models use dVAE tokenization with an 8k-token vocabulary to process images and video frames. Each frame is resized and tokenized separately, resulting in sequences of 256 tokens. These tokens are then processed by a causal transformer that uses the features of RMSNorm and RoPE embeddings to establish improved model performance. The training was done on ImageNet and HowTo100M datasets, tokenizing at a resolution of 128128 pixels. The researchers also optimized the models for downstream tasks by replacing average pooling with attention pooling to ensure a better quality of representation.The models show good performance across the benchmarks. For ImageNet classification, the largest Toto model achieved a top-1 accuracy of 75.3%, outperforming other generative models like MAE and iGPT. In the Kinetics-400 action recognition task, the models achieve a top-1 accuracy of 74.4%, proving their capability to understand complex temporal dynamics. On the DAVIS dataset for semi-supervised video tracking, the models obtain J&F scores of up to 62.4, thus improving over previous state-of-the-art benchmarks established by DINO and MAE. Moreover, on robotics tasks like object manipulation, Toto models learn much faster and are more sample efficient. For example, the Toto-base model attains a cube-picking real-world task on the Franka robot with an accuracy of 63%. Overall, these are impressive results regarding the versatility and scalability of these proposed models with diverse applications.The work provided significant development in video modeling by addressing redundancy and challenges in tokenization. The researchers successfully showed through unified training on both images and videos, that this form of autoregressive pretraining is generally effective across a range of tasks. Innovative architecture and tokenization strategies provide a baseline for further dense prediction and recognition research. This is one meaningful step toward unlocking the full potential of video modeling in real-world applications.Check out the Paper and Project Page. All credit for this research goes to the researchers of this project. Also,dont forget to follow us onTwitter and join ourTelegram Channel andLinkedIn Group. Dont Forget to join our65k+ ML SubReddit. Nikhil+ postsNikhil is an intern consultant at Marktechpost. He is pursuing an integrated dual degree in Materials at the Indian Institute of Technology, Kharagpur. Nikhil is an AI/ML enthusiast who is always researching applications in fields like biomaterials and biomedical science. With a strong background in Material Science, he is exploring new advancements and creating opportunities to contribute. [Recommended Read] Nebius AI Studio expands with vision models, new language models, embeddings and LoRA (Promoted)