Anthropic, the AI company that has long positioned itself as the industry’s most safety-conscious research lab, is dropping the central commitment of its Responsible Scaling Policy — a 2023 pledge to never train an AI system unless it could guarantee beforehand that its safety measures were adequate. “We didn’t really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead,” chief science officer Jared Kaplan told TIME.

The overhauled policy, approved unanimously by CEO Dario Amodei and Anthropic’s board, instead commits the company to matching or surpassing competitors’ safety efforts and to delaying development only if Anthropic considers itself to be leading the AI race and believes catastrophic risks are significant.

The company also plans to publish detailed “Risk Reports” every three to six months and release “Frontier Safety Roadmaps” laying out future safety goals. Chris Painter, director of policy at the AI evaluation nonprofit METR, who reviewed an early draft, told TIME the shift signals that Anthropic “believes it needs to shift into triage mode with its safety plans, because methods to assess and mitigate risk are not keeping up with the pace of capabilities.”

Read more of this story at Slashdot.

Spotify is rolling out a new feature that’s meant to make transitions in between tracks even smoother. If you’ll recall, the streaming service released the ability to create customized transitions within playlists in August last year. It gave people a way to create uninterrupted progressions and eliminate awkward silences between songs. Now, Premium users will be able to make sure the songs in their playlists flow seamlessly even further by reordering tracks based on their keys and BPM or beats per minute.

The new feature can rearrange playlists with one tap. All paying users have to do is tap Mix on one of their playlists and then tap the Edit button. From there, they can scroll down to find the Smart Reorder option. Tapping Smart Reorder will automatically rearrange songs according to their keys and BPM without users having to do anything else. They just have to click Save so that the change to their playlist takes effect.

Spotify says users have streamed over 220 hours of their mixed playlists since it introduced custom transitions last year. It also listed some of the most popular ones on the platform, including The Weeknd’s Wake Me Up transitioning into After Hours and Flo Rida’s Low into Rihann’s S&M.

This article originally appeared on Engadget at https://www.engadget.com/entertainment/music/spotify-can-reorder-your-playlists-by-bpm-and-key-140000101.html?src=rss

Amazon is offering a new way for Alexa+ users to customize the AI assistant’s communication style. The company has introduced three personalities for Alexa+, so the assistant can adopt an attitude that is Brief, Chill or Sweet. 

The Brief style will be exactly that: no small talk and no extra conversation. Chill is easygoing and seems to be inspired by caricatures of the surfer/stoner type, while the Sweet mode is almost aggressively perky and chipper. In the audio sample provided, when a user asks “Alexa, how’s it going?” the Chill voice responds, “Life’s treating me well – all systems are Zen and the digital universe is spinning in harmony.” In contrast, the Sweet one replies, “Absolutely fantastic! I’m radiating pure joy and ready to make your day incredibly amazing!”

Amazon explained that the three personality styles are based on five metrics: expressiveness, emotional openness, formality, directness and humor. The company may release additional options with different combinations of those sliding scale traits in the future.

For now, users can swap the assistant’s vibe from the Alexa app or with the spoken command, “Alexa, change your personality style.” Both approaches can also be used to swap back to the classic Alexa voice. All three personalities are available now for all Alexa+ customers.

This article originally appeared on Engadget at https://www.engadget.com/ai/amazon-introduces-three-personality-styles-for-alexa-140000602.html?src=rss

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

Getting started with outdoor security cameras can get expensive fast. The refurbished Blink Outdoor 4 (two-pack) with Sync Module 2 keeps things simple and affordable. Right now, this “like-new” bundle is $62.99 on Amazon, down from $164.99, and price trackers show this is its lowest price yet. For context, buying new costs $179.99 for just the cameras without the Sync Module. These units are part of Amazon’s certified refurbishment program, so they’ve been tested and inspected before resale. PCMag gave the Blink Outdoor 4 an “excellent” rating, and our own writer called it a worthwhile entry-level wireless camera.

Like-New Blink Outdoor 4 (2-pack) with Sync Module 2

$62.99
at Amazon

$164.99
Save $102.00

$62.99
at Amazon

$164.99
Save $102.00

Installation is simple and fully wire-free. Each camera runs on two AA lithium batteries that Blink says can last up to two years, depending on usage. You mount it on a porch, fence, or even indoors, connect it to wifi in the app, and you’re set. The cameras record in 1080p, with daytime footage clear enough to recognize faces. Plus, it has a 143-degree field of view, so it captures a wide area of your yard or entryway without needing multiple units for basic coverage. Infrared night vision handles low light, and two-way audio lets you speak to a visitor or delivery driver through the app. Motion alerts are sent straight to your phone, and you can access live view for up to five minutes at a time without a subscription.

Storage options are flexible. The included Sync Module 2 supports local video storage through a USB drive (sold separately) and can manage up to 10 Blink cameras at once. Cloud storage is available if you prefer remote backups. Person detection and extended video history require a Blink subscription, with the Basic plan priced at $40 per year for one device and the Plus plan at $120 per year for unlimited cameras.

The system integrates well with Amazon Alexa and IFTTT, so you can pull up live feeds on an Echo Show or Fire TV easily. It does not support Google Assistant or Apple HomeKit. Still, for someone building a basic, outdoor security camera setup on a budget, this refurbished two-pack offers long battery life, local storage, and easy installation at a low upfront cost.

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

Apple users tend to collect chargers: one for the iPhone, one for the Apple Watch, maybe another for AirPods. The Anker MagSafe MagGo UFO 3-in-1 15W Qi2 Wireless Charging Pad is built to replace that pile with a single foldable stand. It’s currently $39.99 on Woot, down from $89.99, with three days left on the deal or until it sells out. The same charger is selling for $76.99 on Amazon, and price trackers show it has dipped to $55.99 in the past, but not this low. Prime members get free standard shipping, while others pay $6. The box includes the charging pad, a 40W wall adapter, and a 5-foot USB-C to USB-C cable, and this wireless charger comes with a two-year Anker limited warranty.

Anker MagSafe MagGo UFO 3-in-1 15W Qi2 Wireless Compact Foldable Travel Charging Pad

$39.99
at Woot

$89.99
Save $50.00

$39.99
at Woot

$89.99
Save $50.00

This is a Qi2-certified three-in-one wireless charger, designed mainly for Apple gear. It can charge an iPhone, an Apple Watch, and a pair of AirPods at the same time. The main magnetic pad delivers up to 15W to compatible iPhones, including the iPhone 15 lineup, which is double the speed of older 7.5W MagSafe chargers. Anker says it can bring an iPhone 15 Pro to 20% in about 15 minutes. The Apple Watch charger also supports fast charging, getting a Series 9 to around 30% in 22 minutes. In everyday use, that means you can top off your phone and watch while you shower or get ready for work and leave with a noticeable boost.

Part of the appeal is the design: It folds down to roughly the size of a baseball, making it easy to toss in a bag for travel. Open it up, and it works as a small stand that supports Apple’s StandBy mode in landscape, so your iPhone can double as a bedside clock or widget display. This is not a universal solution for Android users—it’s best suited to someone already invested in the Apple ecosystem. At $39.99 with the power adapter included, it is a practical upgrade for an Apple-heavy setup at home or on the road.

Uber is one step closer to going airborne. On Wednesday, the company previewed its air taxi booking service ahead of an expected launch in Dubai later this year. The inaugural Uber Air program will let travelers book Joby Aviation’s electric air taxis through a familiar process in the Uber app.

The experience of booking an air taxi will be much like reserving a four-wheeled Uber. In the app, after entering your destination, Uber Air will appear as an option for eligible routes. The Uber app will book a flight and an Uber Black to pick you up and drop you off at a Joby “vertiport.”

Uber

Joby’s air taxis, built exclusively for city travel, can accommodate up to four passengers and luggage. (Uber says size and weight guidelines will be announced closer to launch.) The interior is about the size of an SUV and has “comfortable seating” with panoramic windows. They can travel up to 200 mph and have a range of up to 100 miles. Four battery packs and a triple-redundant flight computer are onboard for safety purposes.

The air taxis aren’t (yet) autonomous and will each have a human pilot onboard. That would at least suggest high prices. After all, pilots aren’t nearly as cheap as Uber’s legion of independent-contractor drivers. But the company insists its air taxi rides will somehow be around as expensive as an Uber Black trip.

Joby

Dubai is only the beginning of the companies’ plans. The US-based Joby says it’s in the final stage of FAA type certification and hopes to launch service in New York and Los Angeles. Globally, it’s targeting the UK and Japan as well.

As for how realistic a US launch is anytime soon, well, that’s up for debate. On one hand, President Trump signed executive orders last year that would create a pilot program to test such aircraft. But safety and cost considerations may require a grounding of expectations.

Joby

In November, Robert Ditchey, a Los Angeles-based aviation expert and test pilot, told NBC News that he didn’t think air taxi service “was ever going to happen” in American cities. “They’re dangerous,” he warned. “We have had helicopters fail and crash on top of buildings in Los Angeles. We’ve had helicopters fail at takeoff and landing in airports. They’re dangerous not from a fire point of view but in terms of landing on top of people and buildings.” In addition, he warned that air taxis can’t be developed in sufficient numbers to make them economically viable “unless they are subsidized by a government.”

Uber and Joby have partnered since 2019. In 2021, Joby bought the Uber Elevate ride-hailing division, which essentially integrated the companies’ services. Last year, Joby acquired Blade Air Mobility’s passenger business, which could open the door to eventually electrifying Blade’s routes.

The video below shows one of Joby’s air taxis taking a test flight in Dubai.

This article originally appeared on Engadget at https://www.engadget.com/transportation/uber-previews-its-dubai-air-taxi-service-130000603.html?src=rss

The silicon landscape shifted Tuesday as Meta Platforms announced a gargantuan, multi-year partnership with Advanced Micro Devices (AMD). For years, the artificial intelligence explosion has been synonymous with NVIDIA’s green logo, but the tide is turning. Meta’s commitment to deploy up to 6 gigawatts of AMD-powered infrastructure—a scale equivalent to the power output of six nuclear reactors—is more than just a purchase order; it is a declaration of independence. By weaving AMD’s upcoming MI450 architecture and 6th Gen “Venice” EPYC CPUs into the heart of its personal superintelligence roadmap, Meta is systematically dismantling the NVIDIA stranglehold.

The Strategic Divergence: AMD vs. NVIDIA in the Meta Ecosystem

To understand the magnitude of this deal, one must look at how it differs from Meta’s existing relationship with NVIDIA. Last week, Meta doubled down on NVIDIA’s Blackwell and Rubin platforms in a deal valued at approximately $50 billion. That partnership is a “full-stack” marriage, where Meta buys into NVIDIA’s proprietary vision, including the CUDA software layer and InfiniBand networking. It is the premium, “it just works” option for the most grueling frontier model training, like Llama 4.In contrast, the AMD deal is a co-engineering masterclass. Meta isn’t just buying off-the-shelf parts; they are collaborating on a customized MI450-based GPU optimized specifically for Meta’s unique inference workloads. While NVIDIA provides the “black box” of high-performance computing, AMD is providing the “open toolkit.” This deal centers on the Helios rack-scale architecture, an Open Compute Project (OCP) compliant design that emphasizes modularity and energy efficiency. Where NVIDIA offers a locked ecosystem, AMD offers a collaborative roadmap that allows Meta to vertically integrate its software stack, ROCm, directly into the silicon.

Advancing the AMD Roadmap: From Challenger to Peer

This partnership is the ultimate validation of Lisa Su’s “AI Everywhere” initiative. For years, the critique of AMD was that its hardware was capable, but its software ecosystem was immature. By securing Meta as a lead partner for the MI400 series, AMD has gained a massive live-fire testing ground to refine its software stack.

The deal effectively pulls forward AMD’s transition into the “Tier-1” hyperscaler group. It provides the necessary capital and telemetry data to accelerate the development of unreleased products. Specifically, the MI455X and the upcoming MI500 series will benefit from the architectural feedback generated by Meta’s gigawatt-scale deployment. By the time these chips reach the broader market, they will have been “hardened” by one of the world’s most demanding AI environments, making them far more attractive to other cloud providers who were previously hesitant to stray from the NVIDIA path.

The Art of the Squeeze: Why Two Giants Are Better Than One

The most brilliant aspect of Meta’s strategy is the creation of a perfectly competitive duopoly. By committing tens of billions of dollars to both camps, Mark Zuckerberg has turned the AI chip market into a buyer’s market.

When NVIDIA was the only game in town, they held all the cards—pricing power, allocation priority, and roadmap control. By bringing AMD to parity in its data centers, Meta can now play the two giants against each other. If NVIDIA tries to hike margins on Blackwell, Meta can shift more of its 6-gigawatt expansion toward AMD’s Helios racks. If AMD lags on software optimization, Meta can pivot back to NVIDIA’s “Vera” CPUs. This dual-sourcing strategy is estimated to give Meta the leverage to negotiate prices up to 30% lower than a single-source customer could ever dream of, effectively saving billions in capital expenditure over the next five years.

Unreleased Power: The Impact on MI450 and “Venice”

A significant portion of this deal hinges on hardware that the public hasn’t even touched yet. The AMD Instinct MI450, built on an advanced 3nm process, is the centerpiece. This chip is expected to offer a massive leap in HBM4 memory bandwidth and FP4/FP8 compute performance. Because Meta is a “definition customer,” the MI450 is being tuned to handle the massive transformer-based models that power Instagram and WhatsApp.

Furthermore, the deal secures Meta as the lead customer for the 6th Generation EPYC “Venice” CPUs. These processors are designed to eliminate the “CPU bottleneck” often found in AI clusters, ensuring that the GPUs are never waiting for data. The synergy between the Venice CPUs and the MI450 GPUs, linked by Infinity Fabric, creates a unified compute fabric that NVIDIA’s proprietary NVLink struggles to match in terms of open-standard flexibility.

The Responsiveness Gap: Why Meta is Looking Elsewhere

There is a growing sentiment in the industry that NVIDIA has become a victim of its own success. As the undisputed king of AI, NVIDIA has earned a reputation for being “difficult” to work with—dictating terms and forcing adoption of its entire hardware/software stack.

Meta, which prides itself on its Open Compute Project heritage, finds this “walled garden” approach restrictive. AMD, in its underdog role, has been far more willing to listen. AMD’s engineers have worked side-by-side with Meta to co-develop the Helios rack and optimize the ROCm stack for Meta’s environment. This partnership sends a clear signal to Santa Clara: if NVIDIA does not become more responsive to customer needs and more flexible with its licensing and integration, it faces total displacement in the world’s largest data centers. The threat of losing a 6-gigawatt customer is the only thing powerful enough to force NVIDIA to lower its barriers.

Wrapping Up

The AMD-Meta partnership is a watershed moment for the technology industry. It effectively ends the era of the AI monopoly and ushers in an age of strategic multi-sourcing. For AMD, it provides the scale and validation needed to compete at the absolute highest level of the data center. For Meta, it provides the economic leverage to fuel its “personal superintelligence” vision without being held hostage by a single supplier’s pricing. As the first gigawatts of MI450-powered servers go online, the industry will be watching closely. If AMD can match NVIDIA’s performance at scale, the green giant may finally have to start playing by someone else’s rules.

HP has revealed that memory now accounts for 35% of the cost of materials it needs to build a PC, up from between 15 and 18% last quarter. And the company expects RAM’s contribution will rise through the year. From a report: Speaking on the company’s Q1 2026 earnings call, interim CEO Bruce Broussard said the company has secured long-term supply agreements for the year and also “qualified new suppliers [and] built in strategic inventory positions for key platforms and cut the time to qualify new material in half to accelerate our product configuration changes.”

That sounds a lot like HP Inc is signing up new suppliers at a brisk pace. Broussard said the company has also “expanded lower-cost sourcing across our commodity basket, lowering logistics costs with agile end-to-end planning processes.” The company is using its internal AI initiatives to power those new processes. The company is also “configuring our products and shaping demand to align the supply we have with our customer needs” and “taking targeted pricing actions to offset the remaining cost impact in close partnership with both our channel and direct customers.”

Read more of this story at Slashdot.

Ahead of a full release at Mobile World Conference (MWC), Honor has teased the MagicPad 4 that it calls the world’s thinnest Android tablet. The new model is just 4.8mm thick (not counting that camera bump), a full millimeter thinner than the MagicPad 3 and slightly less than the 5.1mm iPad Pro and Samsung Galaxy Tab S11, the company revealed.

On top of being thinner, the MagicPad 4 has a new 12.3-inch 165Hz OLED display. While slightly smaller than before, it should be considerably better than the LCD display on the previous model. The new model weighs 145 grams less than before at 450g thanks to that screen and the slightly smaller 10,100 mAh battery (with a 66W fast charger in the box). 

The new tablet is powered by Qualcomm’s Snapdragon 8 Gen 5 chipset and comes with up to 16GB of RAM and 512GB of storage. It’s equipped with 13MP rear and 9MP front cameras, along with eight speakers for spatial audio. The MagicPad 4 will run MagicOS 10, Honor’s flavor of Android 16. There’s no word on pricing or availability yet, but we’ll likely learn more at the company’s press conference on Sunday — along with the company’s weird robot revealed yesterday. 

This article originally appeared on Engadget at https://www.engadget.com/mobile/tablets/honor-says-its-48mm-thick-magicpad-4-is-the-worlds-slimmest-android-tablet-114346615.html?src=rss

OpenAI has successfully convinced the court to dismiss the lawsuit filed by Elon Musk’s xAI, accusing the company of stealing its trade secrets. In her decision, US District Judge Rita F. Lin wrote that xAI’s complaint “does not point to any misconduct by OpenAI” and instead attributes all listed misconducts to its eight former employees who “ left for OpenAI at around the same time.”

Lin said that xAI accused two of its former employees of stealing its source code before leaving at a time when they were already speaking to an OpenAI recruiter. However, the company didn’t say if the recruiter told those former employees to do so. xAI’s lawsuit also accuses two other former employees of keeping their work chats on their devices even after leaving, another of refusing to provide certifications related to confidential information after his departure, and another of unsuccessfully trying to access xAI hiring and datacenter optimization information when he was already working for OpenAI.

“Notably absent are allegations about the conduct of OpenAI itself,” the judge noted. xAI didn’t include any information that directly accuses OpenAI of making those employees steal its trade secrets. It also didn’t include allegations that those former employees used any stolen trade secrets after they were already working for OpenAI. To be precise, OpenAI’s motion for dismissal was granted with leave to amend, so the lawsuit may not be completely over just yet. That means xAI can still file an amended complaint addressing what the judge wrote in her decision until March 17, 2026.

OpenAI and xAI have a longstanding feud, and this is just one of the several lawsuits between the two companies. In fact, Musk has an ongoing complaint against OpenAI and Microsoft, accusing the former of violating its nonprofit status. Musk, who was an early funder of OpenAI, is now asking the company for $79 billion to $134 billion in damages from “wrongful gains.”

This article originally appeared on Engadget at https://www.engadget.com/ai/xais-trade-secret-lawsuit-against-openai-has-been-dismissed-101912599.html?src=rss