Posts with «author_name|richard lai» label

Xiaomi 14 Ultra combines a 1-inch camera sensor with four AI imaging models

Xiaomi just couldn't wait until MWC to unveil its latest Leica-endorsed flagship phone. Following the 12S Ultra and 13 Ultra, Xiaomi is finally catching up with the competition by picking up Sony's second-gen 1-inch mobile camera sensor, the LYT-900, for its brand new 14 Ultra flagship phone. This marks the second device to don this crème de la crème of imaging silicons, after Oppo's Hasselblad-tuned Find X7 Ultra from early January. That said, the Xiaomi 14 Ultra does have a slight edge with its faster main variable aperture at up to f/1.63, beating the Find X7 Ultra's f/1.8 — on paper, at least.

With the exception of the faster f/2.5 aperture on the new 120mm periscope shooter, the remaining three Summilux rear cameras are almost identical to the previous set, and they are still powered by a Sony IMX858 sensor each. The biggest change in this field is the new Xiaomi AISP neural chip, which claims to be the world's first AI large-model computational platform for photography. This leverages four large models — "FusionLM," "ToneLM," "ColorLM" and "PortraitLM" — to fine-tune results, especially with digital zoom at 30x or more.


The 14 Ultra also packs some surprises in the battery, durability and connectivity categories. As seen in the super-slim Mix Fold 3 and Honor Magic V2, the 14 Ultra is Xiaomi's first candybar to jump onto the silicon carbon cell bandwagon, in order to pack 5,300mAh of juice into a space that's 8 percent smaller. Xiaomi claims that compared to the previous model, you get 17-percent more stamina with this battery upgrade. To replenish the battery, you get both 90W of wired fast charging and 80W of wireless fast charging — these take 12.5 minutes and 20 minutes to reach 50 percent, respectively.

Going along with the "Year of the Dragon" theme, Xiaomi claims that the 14 Ultra's "Dragon Armor" structure has double the bending resistance, thanks to its special "6M42" aluminum alloy mid-frame (supposedly crafted with a better grip as well). The Chinese brand even claims that this part is 8-percent stronger than the iPhone 15 Pro's titanium frame." aluminum alloy mid-frame (supposedly crafted with a better grip as well). The Chinese brand even claims that this part is 8-percent stronger than the iPhone 15 Pro's titanium frame, but it decided to offer a more premium titanium version as well.


This metallic frame is complemented by a "Dragon Crystal" glass — shielding the 6.73-inch AMOLED screen (3,200 x 1,440 120Hz; made by TCL CSOT) — with apparently 10 times more drop resistance. Xiaomi also touts its new vegan leather material, which has been certified by SGS to have six times more wear resistance, has more dirt resistance and is less prone to yellowing due to ultraviolet rays — an important breakthrough particularly for the white version. But if you prefer something shiny, the 14 Ultra is also available in a blue "Dragon Crystal" ceramic flavor, which resembles ceramic but isn't as heavy — it only weighs 5 grams more than its vegan leather counterpart. Regardless of the cover material, this device has IP68 rating for dust and water resistance.

Much like the 14 and 14 Pro from October (and the SU7 electric sedan's in-car entertainment system), the 14 Ultra runs on Xiaomi's Android-based HyperOS, and it's powered by Qualcomm's latest Snapdragon 8 Gen 3 processor. This is cooled by a dual-loop vapor chamber, which also sucks heat out of the camera modules. The processor is backed by Xiaomi's new proprietary chip, the Surge T1, which apparently boosts cellular connectivity by up to 37 percent, as well as Wi-Fi and Bluetooth connections by up to 16 percent.


This device also supports two-way satellite calling and texting, now with 60-percent faster satellite locking and 29-percent faster satellite connection. As a bonus, when you're lost, you can send your location data along with vital signs from your wearable device — presumably exclusive to one of the latest Xiaomi watches or smart bands. Sadly, these satellite features are likely limited to China for now.

We'll likely be hearing about the Xiaomi 14 Ultra's global launch at MWC next week, but for now, we can refer to the Chinese pre-order pricing. The vegan leather and ceramic variants all start from 6,499 yuan (about $900) for the 12GB of RAM and 256GB of storage configuration, and max out at 7,799 yuan ($1,080) with 16GB of RAM and 1TB of storage. These will be available for retail from February 27. The titanium version with dark gray vegan leather is based on the top configuration but costs an extra 1,000 yuan ($140), and it won't be available until March 12.


Likes its predecessor, the 14 Ultra has an optional photography kit with a shutter button grip that adds an extra 1,500mAh of power. The upgrade this time is the new video recording button, along with a customizable jog dial. You can get this accessory for 699 yuan ($100) as a bundle with the phone.

This article originally appeared on Engadget at

Microsoft's upcoming custom chip will be made by Intel

Intel's relatively new Foundry division — formerly known as Intel Foundry Services until earlier today — has just landed a notable order from a big name. According to Bloomberg and The Wall Street Journal, Microsoft CEO Satya Nadella announced that his company will be tapping into Intel's latest 18A (1.8nm) fabrication process for an upcoming in-house chip design. But given Intel's process roadmap, this means we likely won't be seeing Microsoft's new chip until 2025.

While neither company disclosed the nature of said silicon, Microsoft did unveil its custom-made Azure Maia AI Accelerator and Azure Cobalt 100 CPU server chips last November, with an expected rollout some time "early" this year to bolster its own AI services. The Cobalt 100 is based on Arm architecture, and it just so happens that Intel has been optimizing its 18A process for Arm designs since April last year (it even became an Arm investor later), so there's a good chance that this collaboration may lead to the next-gen Cobalt CPU.

In addition to the usual efficiency improvements as node size decreases, Intel 18A also offers "the industry's first backside power solution" which, according to IEEE's Spectrum, separates the power interconnect layer from the data interconnect layer at the top, and moves it to beneath the silicon substrate. This allows for improved voltage regulation and lower resistance, which in turn enable faster logic and lower power consumption, especially when applied to 3D stacking.

In Intel's Q4 earnings call, CEO Pat Gelsinger confirmed that "18A is expected to achieve manufacturing readiness in second half '24." Given that Intel's very own 18A-based processors won't arrive until 2025, chances are it'll be a similar time frame for Microsoft's next chip.


At Intel's event earlier today, the exec shared an extended Intel Foundry process technology roadmap, which features a new 14A (1.4nm) node enabled by ASML's "High-NA EUV" (high-numerical aperture extreme ultraviolet) lithography system. According to AnandTech, this 14A leap may help Intel play catchup after its late EUV adoption for its Intel 4 (7nm) node, though risk production won't take place until the end of 2026.

Intel Foundry is the brainchild of Gelsinger, who launched this department right after he assumed the CEO role in February 2021, as part of his ambitious plan to put Intel up against the likes of TSMC and Samsung in the contract chip-making market. Before Microsoft, Intel Foundry's list of clients already include MediaTek, Qualcomm and Amazon. The company still aims to become "the second largest external foundry by 2030" in terms of manufacturing revenue, which it believes is achievable as early as this year.

This article originally appeared on Engadget at

Uber Eats expands its autonomous food delivery service to Japan

Following its autonomous food delivery launch in Miami and Fairfax, Virginia, Uber Eats will soon be offering the same robotic service in Japan — its first outside the US. It is once again collaborating with Google alum startup Cartken, with local compliance help from Mitsubishi Electric, to bring a fleet of Model C sidewalk delivery robots to select areas in Tokyo in March. Uber Eats Japan CEO Shintaro Nakagawa says the autonomous delivery service will solve the local labor shortage issue, while complementing the existing human delivery methods "by bicycle, motorbike, light cargo, and on foot."

Cartken's six-wheeled Model C uses six cameras and advanced AI models for autonomous driving plus obstacle detection, and remote control mode is available when needed. With guidance from Mitsubishi, the robot has been modified to suit local needs in Japan. For one, its speed is capped at 5.4 km/h or about 3.36 mph as per local regulation, which is a lot slower than the 6 mph top speed it's actually capable of. The loading capacity has also been reduced from 1.5 cubic feet to about 0.95 cubic feet (27 liters), likely due to the extra thermal insulation in the compartment. Uber Eats adds that for the sake of privacy, people's faces are automatically masked in footage captured by the robots.

While this is Uber Eats' robotic delivery debut in Japan, Cartken already has a presence there thanks to Mitsubishi. Since early 2022, the duo has worked with Starbucks, local e-commerce giant Rakuten and supermarket chain Seiyu in some parts of Japan. In the US, Cartken also has a partnership with Grubhub to provide autonomous food delivery service on college campuses, including the Ohio State University and the University of Arizona.

Even though Uber Eats has yet to share which Tokyo restaurants will be tapping into its robotic delivery service, it should have no problem seeking partnership given Cartken's prior local experience. That said, I highly doubt that the pair would risk trialing their robots through a crowd of drunkards in Shibuya just yet.

This article originally appeared on Engadget at

Signal usernames will keep your phone number private

Instant messaging app Signal is best known for its privacy-related settings, though with phone numbers being the heart of the platform since its inception, there was no way to fully hide your own number until now. Earlier today, Signal announced that you'll soon be able to create a unique username (not to be mistaken with your profile name), which you can share with others via a link or QR code — as opposed to sharing your number. You'll be able to change your unique username as often as you want, but it needs to contain two or more numbers at the end, as part of Signal's anti-spoofing efforts. You can even delete your username entirely, as it is an optional feature.

Naturally, you'll still need a phone number to sign up for Signal, but note that with the new default, your number will no longer be visible to everyone (you can change this setting manually, if needed). There will also be a new option which prevents people from finding you by your number; they will need to have your exact unique username to do so. In other words, people who already have your number won't necessarily know that you are also on Signal, which is a good thing if you prefer to stay anonymous in this platform's public groups.

As is the case with any new feature, the likes of spammers and scammers will eventually find a way to abuse usernames, as you won't be able to verify their numbers instantly. Pro tip: when you see new contacts that appear to be your acquaintances, always double check with them through other means — preferably in person, or at least via a phone call. You may look out for these new Signal features in a few weeks' time, or you can get an early taste in the beta release.

This article originally appeared on Engadget at

The world's thinnest foldable phone gets a Porsche Design makeover

When Honor's Magic V2 first showed up some seven months ago, we were left impressed with its sub-1cm folded thickness. The company has since picked up where Huawei left off and joined forces with Porsche Design, with the first result being a special edition foldable phone, the "Magic V2 RSR" ("RSR" stands for "RennSport Rennwagen," which translates to "racing sport racing car"). Porsche fans may recognize several attributes here that pay homage to the automotive brand, including the iconic agate gray along with a flyline running down the middle of the body — supposedly resembling the Porsche 911's hood. The rear camera island has also been upgraded with a dark titanium frame to go with its bolder trapezoidal redesign.

The Magic V2 RSR retains the same dimensions, measuring 9.9mm thick when folded and 4.7mm when opened, ensuring it remains the world's thinnest foldable phone. The extra decoration adds 3 grams over the base model, reaching 234 grams (about 8.25 oz) which still beats the likes of the OnePlus Open, Samsung's Galaxy Z Fold 5 and Google's Pixel Fold, largely thanks to Honor's mix of titanium alloy and proprietary steel in its hinge design — this is apparently certified to last for at least 400,000 folds. As a bonus, the Magic V2 line's 5,000mAh battery capacity is larger than the competition, thanks to its more energy-dense silicon-carbon cells. The trade-off here is the missing wireless charging, but you do get 66W of wired fast charging, which takes around 50 minutes to go from zero to 100 percent.

The Magic V2 RSR is still powered by Qualcomm's Snapdragon 8 Gen 2 processor, but with storage option limited to just 1TB to go with the 16GB of RAM. The rest of the spec sheet is identical for both Magic V2 variants. For displays, you get a 7.92-inch 2,344 x 2,156 foldable OLED screen (with a barely-noticeable crease) and a 6.43-inch 2,376 x 1,060 exterior OLED display, with their brightness going up to 1,600 nits and 2,500 nits, respectively. For audio, this Android phone comes equipped with "IMAX Enhanced" stereo speakers, whereas phone calls leverage the device's three-mic system for better voice reception. As is the case with most foldables, the fingerprint reader is integrated into the side-mounted power button.

The rear "Falcon Camera System" consists of a 50-megapixel main sensor, a 50-megapixel ultra-wide shooter and a 20-megapixel telephoto camera (2.5x zoom), whereas a 16-megapixel punch-hole selfie camera sits at the top of both screens. Both the rear and front cameras support up to 4K resolution for video recording.

It should come as no surprise that the Magic V2 RSR boxset packs some extra goodies, namely an active stylus, its own carrying case and a spare 66W charger. The special protection case is wrapped in a stitched dark gray vegan leather, though compared to the normal version, it's missing out on a kickstand — a feature I struggle to live without when it comes to foldable phones.

Photo by Richard Lai / Engadget

Like the base model, the Magic V2 RSR will also be headed to global markets, though the prices are under wraps until MWC later this month. For reference, the Magic V2 RSR retails for 15,999 yuan or about $2,220 in China, but expect a huge bump for international pricing. The regular Magic V2 — only available in 512GB flavor for international markets — is priced at £1,700 ($2,140) in the UK and €1,999 ($2,150) in Europe, but the China version and Hong Kong version (also international firmware) cost around $1,400 only. It's no wonder that some savvy consumers would rather import smartphones from overseas, despite the lack of local warranty services.

This article originally appeared on Engadget at

New York City is suing social media firms for allegedly harming the mental health of children

After designating social media as a "public health hazard" in late January, New York City is now suing Meta, Google, Snap and TikTok for "fueling nationwide youth mental health crisis." Specifically, these companies face three counts in the lawsuit: public nuisance, negligence and gross negligence. The Mayor Eric Adams administration accuses TikTok, Instagram, Facebook, Snapchat and YouTube of "endangering our children's mental health, promoting addiction, and encouraging unsafe behavior."

These are allegedly achieved by way of harmful algorithms, gambling-like mechanisms and manipulation through reciprocity — making the user "feel compelled to respond to one positive action with another positive action." The city believes that there is a correlation between the increase in social media usage and the decline in local youth mental health over "more than a decade."

In response, Google and Meta told CNBC that they have always worked with youth safety experts and provided parental control tools. ByteDance's TikTok also highlighted some of its specific tools to Axios, namely age-restricted features, parental controls and an automatic 60-minute time limit for users under 18. However, none of the tech companies acknowledged the problematic features listed by the Adams administration.

This lawsuit follows a recent Senate hearing on online child safety, in which the CEOs of all the aforementioned tech companies (except Google) were present. In his opening remarks, Senator Lindsey Graham told the tech execs that "you have blood on your hands" — a reference to online child exploitations and cyberbullying that unfortunately led to deaths. 

Through this case, the Adams administration wants these tech companies to pay up for the city's youth mental health services, which apparently cost more than $100 million each year. But ultimately, it's about forcing these tech giants to stop manipulating young users into addictive behavior, as well as getting policymakers to place new federal laws that safeguard youth mental health on social platforms.

Before this New York City lawsuit, Meta already faces a similar case from 41 states back in October 2023, in which it was accused of misleading the public about the safety of its platform's "addictive" features. Meta, Snap, TikTok and Google were also sued in a multi-district litigation in 2022 for their addictive features that allegedly cause "emotional and physical harms, including death" to adolescents.

This article originally appeared on Engadget at

Dyson's new lightweight 'Supersonic r' hairdryer looks a lot like a periscope

Dyson's signature Supersonic hair dryer hasn't changed much since its launch back in 2016, so the brand caught us by surprise when it unveiled a brand new model with a completely different look at New York Fashion Week earlier today. The "Supersonic r" is a tube-shaped hair dryer that dwarfs its predecessor, and it's lighter as well, weighing just 325g (around 11.5 ounces) — almost half that of the original Supersonic, which my hairstylist friend said was too heavy for all-day use. 

This impressive weight and size reduction is largely thanks to some re-engineered heating elements, which allowed the team to replace the old design's large head with just a bend at the end of the shaft — like a lower case "r," hence the name. (Remember the HTC RE camera?) Miniaturized heating grills consisting of 99.9% copper microfins are lined up along the pipe's curve, to ensure the high-pressure airflow is heated evenly with high precision (via intelligent heat control) to achieve better hair shine. The elements are also laid out in a way that avoids hot spots being generated, which may otherwise cause heat damage on your hair.


The buttons for three precise airflow settings and four heat modes (including constant cold shot) are just a thumb-click away. The air inlet at the bottom of the handle now comes with a longer-lasting depth-loaded filter, which captures air pollutants typically present in hair salons.

In addition to Dyson's improved Hyperdymium motor in the handle, the Supersonic r is also equipped with an RFID sensor to recognize its magnetic accessories — diffuser, flyaway attachment, wide tooth comb, pro concentrator and a new powerful air attachment. This feature allows the hair dryer to automatically adjust the motor and heater accordingly, in order to deliver optimal airflow and temperature to suit the purpose. On example is that the diffuser is set to low flow and low temperature, which reduces frizz and enhance curls. Each attachment can also save your customized settings for future use.


Dyson's Supersonic r is asking for $570, but it's currently aimed at professionals, meaning us mere mortals will have to figure out other ways to get hold of one. If you happen to be a US-based hairstylist with a valid cosmetology license, feel free to join the waitlist, and your salon can pick up a Supersonic r as soon as April.

This article originally appeared on Engadget at

ChatGPT will digitally tag images generated by DALL-E 3 to help battle misinformation

In an age where fraudsters are using generative AI to scam money or tarnish one's reputation, tech firms are coming up with methods to help users verify content — at least still images, to begin with. As teased in its 2024 misinformation strategy, OpenAI is now including provenance metadata in images generated with ChatGPT on the web and DALL-E 3 API, with their mobile counterparts receiving the same upgrade by February 12.

The metadata follows the C2PA (Coalition for Content Provenance and Authenticity) open standard, and when one such image is uploaded to the Content Credentials Verify tool, you'll be able to trace its provenance lineage. For instance, an image generated using ChatGPT will show an initial metadata manifest indicating its DALL-E 3 API origin, followed by a second metadata manifest showing that it surfaced in ChatGPT.

Despite the fancy cryptographic tech behind the C2PA standard, this verification method only works when the metadata is intact; the tool is of no use if you upload an AI-generated image sans metadata — as is the case with any screenshot or uploaded image on social media. Unsurprisingly, the current sample images on the official DALL-E 3 page returned blank as well. On its FAQ page, OpenAI admits that this isn't a silver bullet to addressing the misinformation war, but it believes that the key is to encourage users to actively look for such signals.

While OpenAI's latest effort on thwarting fake content is currently limited to still images, Google's DeepMind already has SynthID for digitally watermarking both images and audio generated by AI. Meanwhile, Meta has been testing invisible watermarking via its AI image generator, which may be less prone to tampering.

This article originally appeared on Engadget at

Roblox adds real-time AI chat translation using its own language model

Currently serving over 70 million daily active users, Roblox is still going strong since its September 2006 launch — almost 18 years ago. The development team is now taking one step further to boost the platform's massive community, by way of providing real-time AI chat translation to connect gamers around the world. According to CTO Daniel Sturman, his team needed to build their own "unified, transformer-based translation LLM (large language model)" in order to seamlessly handle all 16 languages supported on Roblox, as well as to recognize Roblox-specific slangs and abbreviations (this writer just learned that "obby" refers to an obstacle course in the game).

As a result, the chat window always displays the conversation in the user's own tongue — with a small latency of around 100 milliseconds, so it's pretty much real time. You can also click on the translation icon on the left of each line to see it in its original language. Sturman claims that thanks to the language model's efficient architecture and iterative training, it "outperforms commercial translation APIs on Roblox content." The development team will later roll out a feedback tool to help improve translation quality, in addition to its ongoing updates with whatever new catchphrases it picks up on the platform.


Roblox's translation efforts don't stop there. Sturman adds that his team is already looking into automatically translating "text on images, textures, 3D models" and more. As Roblox supports voice chat, the exec also teases the possibility of automatic voice chat translations, so gamers from around the world can seamlessly talk to one another in their own tongue on the platform. Given that Samsung already offers a similar feature via Galaxy AI, it probably won't be long before we hear another update from Roblox on this end.

This article originally appeared on Engadget at

Apple Vision Pro teardown deconstructs the weird-looking EyeSight display

One of Vision Pro's most intriguing features is undoubtedly the EyeSight display, which projects a visual feed of your own eyes to better connect with people in the real world — because eye contact matters, be it real or virtual. As iFixit discovered in its teardown, it turns out that Apple leveraged stereoscopic 3D effect as an attempt to make your virtual eyes look more life-like, as opposed to a conventional "flat" output on the curved OLED panel. This is achieved by stacking a widening optical layer and a lenticular lens layer over the OLED screen, which is why exposing the panel will show "some very oddly pinched eyes." The optical nature of the added layers also explain the EyeSight display's dim output. Feel free to check out the scientific details in the article.

While iFixit has yet to do more analysis before it can give the Vision Pro a repairability score, so far we already know that the front glass panel "took a lot of heat and time" to detach from the main body. That said, the overall modular design — especially the speakers and the external battery — should win some points. As always, head over to iFixit for some lovely close-up shots of the teardown process.

This article originally appeared on Engadget at