AI-powered visual search features arrived to Ray-Ban's Meta sunglasses last year with some impressive (and worrying) capabilities — but a new one in the latest beta looks quite useful. It identifies landmarks in various locations and tells you more about them, acting as a sort of tour guide for travelers, Meta CTO Andrew Bosworth wrote in a Threads post. 

Bosworth showed off a couple of sample images explaining why the Golden Gate Bridge is orange (easier to see in fog), a history of the "painted ladies" houses in San Francisco and more. For those, the descriptions appeared as text below the images.

On top of that, Mark Zuckerberg used Instagram to show off the new capabilities via a few videos taken in Montana. This time, the glasses use audio to provide a verbal description of Big Sky Mountain and the history of the Roosevelt Arch, while explaining (like a caveman) how snow is formed. 

Meta previewed the feature at its Connect event last year, as part of new "multimodal" capabilities that allow it to answer questions based on your environment. That in turn was enabled when all of Meta's smart glasses gained access to real-time info (rather than having a 2022 knowledge cutoff as before), powered in part by Bing Search.

The feature is part of Meta's Google Lens-like feature that enables users to “show” things they are seeing through the glasses and ask the AI questions about it — like fruits or foreign text that needs translation. It's available to anyone in Meta's early access program, which is still limited in numbers. "For those who still don’t have access to the beta, you can add yourself to the waitlist while we work to make this available to more people," Bosworth said in the post. 

Despite combining for 32 nominations, Netflix and Apple TV+ were nearly shut out of of the 2024 Oscars, with Netflix winning a single award for Wes Anderson's The Wonderful Story of Henry Sugar (Best Live Action Short Film). The big surprise was Martin Scorsese's Killers of the Flower Moon (Apple TV+) not gaining a single statue, notably Lily Gladstone losing the Best Actress prize to Poor Things' Emma Stone. 

Universal was the big winner with Oppenheimer (Best Picture, Best Director, editing, Best Actor, Best Supporting Actor, original score, cinematography) while Poor Things (Disney's Searchlight Pictures) garnered four prizes. The other half of "Barbenheimer," (Barbie, ofc) took just a single prize for best song with What Was I Made For (and not I'm Just Ken, sung live at the ceremony by supporting actor nominee Ryan Gosling). 

The gala was a letdown for Netflix, which scored six prizes last year. Netflix has won 23 Oscars since 2017, but has yet to win in the Best Picture or Best Actor/Actress categories. That's despite four nominations this year for Bradley Cooper and Carey Mulligan (Maestro), Colman Domingo for Rustin and Annette Bening for Nyad. 

Lily Gladstone was perhaps slightly favored over Stone to win for Killers of the Flower Moon and she also would have been the first Indigenous American to win an Academy Award for Best Actor or Actress. The studio was also up for Best Supporting Actor with nominee Robert De Niro and Best Director for Scorsese. Apart from Gladstone's loss, though, awards were distributed largely as anticipated. 

This year, it couldn't be said that judges were swayed by a lack of theatrical presence from streamers. Killers of the Flower Moon had a fairly wide release in cinemas, while Maestro stayed in theaters for a month prior to its Netflix release. Both chalked up decent box office numbers. 

The ceremony itself appears to have been decent, with Deadline proclaiming that producers "finally made an Academy Awards ceremony for the 21st century" and The Hollywood Reporter calling it "busy and eclectic." USA Today did describe it as "boring," but every Oscars for the past 20 years has gotten the same knock.

Rivian generally had a good day yesterday, launching the R2 SUV along with the surprise R3 crossover and dune buggy-esque R3X that were met with general acclaim. Buried in that press release, however, was the news that the automaker is halting production of its $5 billion Georgia plant in order to save money. 

Instead of building the R2 in Georgia as originally planned, the company will start production of the electric SUV at its existing Normal, Illinois plant. "Beyond significantly reducing the amount of capital needed to bring R2 to market, the company believes this approach considerably reduces risk to the launch and associated ram," the company said. 

The move will also allow Rivian to bring the R2 to market sooner, in the first half of 2026, while saving the company $2.25 billion in capital spending in the short term. That's important since it has been burning through cash of late, according to recent reports.

Elliot Ross Studio

Of all the EV startups to come along of late, Rivian has been one of the most promising thanks to significant investments from Amazon, Ford and others. The company's electric R1T pickup and R1S SUV were also widely praised for their attractive designs, healthy range and more. 

Ramping up an automotive startup is no easy feat, though, especially in a market that's been tough on EVs of late — with even stalwart Tesla feeling the pinch. It doesn't help that startup rivals like Fisker are having serious cash flow issues, as it may spook consumers wary of untested EV brands.

Rivian selected Georgia as the site for its second EV factory back in 2021, receiving up to $1.5 billion in state incentives. At the time, the company said it hoped to eventually produce 400,000 electric vehicles there annually. With plant changes, the Normal, Illinois facility will augment capacity to 215,000 units annually across R1T, R1S, EDV, RCV, and R2.

The Georgia location remains in the picture, but Rivian only said it construction would restart later. "Rivian’s Georgia plant remains an extremely important part of its strategy to scale production of R2 and R3," it said in a statement. 

Leica's SL family has always been the most "mirrorless" looking of its cameras, offering pro photographers handling that's similar to rivals like Canon and Sony. That continues with the launch of the SL3, a 60-megapixel (MP) model that looks nothing like its famous rangefinder cameras, offering a new phase-detect autofocus system, 8K video capability and more.

The new model has the same sensor as the M11 rangefinder and Q3 compact models, but very much resembles the SL2 and tips the scale at a hefty 767 grams. It's also IP54-rated so you can shoot in weather conditions ranging from 14 to 104°F without worrying about dust, moisture etc. 

Leica

The biggest change in the layout is a new dial on the left top (all the dials are blank as you can customize them any way you want). At the back, Leica moved the three key buttons (play, function and menu) to the right, and changed the power switch to a button. 

The SL3 now has a new 3.2″ 2.3m-dot LCD, and unlike past models, it can tilt (but not swivel). The EVF has also been updated to a higher-resolution 5.76m-dot 0.78x OLED version. As for storage, it supports both CFexpress type B and SD UHS-II memory cards. It comes with a full-sized HDMI and USB-C charging ports, along with mic/headphone jacks.

Leica

The new sensor increases resolution from 47 MP on the SL2 up to 60 MP, a significant bump. It's likely similar to the sensor Sony uses on the A7R V, and supports a wide native ISO range from 50-100,000, with a claimed 15 stops of dynamic range. 

It can shoot bursts up to 15fps, but more importantly, offers a new phase-detect AF system that's faster and more intelligent, according to the company. It combines phase detection, contrast and object autofocus for faster and more consistent performances. That includes face and eye detection, along with a new animal detection feature. It's also supposed to be better at tracking action, Leica says. 

Leica

As for video, the SL3 can record internally in DCI 8K 8192 x 4320 at 30 fps and DCI 4K at up to 60p with a maximum bitrate of 600Mb/s. It can also record externally via the HDMI output, also at DCI 8K and DCI 4K up to 60fps. Those resolutions are all available in 10-bit with the choice of L-Log Rec. 2020 and HLG Rec. 2020. It supports ProRes recording, but only up to 1080p. 

As you're probably expecting, the SL3 isn't cheap at $6,995, though it is significantly cheaper than the $9,000 M11. It's now available at B&H Photo Video and elsewhere.

Sony is so closely associated with full-frame mirrorless cameras that it’s easy to forget it also sells the A6000 APS-C lineup — particularly since the last one, the A6600, came out five long years ago in 2019. The flagship A6700 finally arrived last year, though, with a feature list that appeared to be worth the wait.

It’s the same price the A6600 was at launch four years ago, but addresses its predecessor’s main flaws by boosting resolution a bit and reducing rolling shutter. At the same time, it’s been likened to a mini-FX30 cinema camera as it has the same sensor and video capabilities.

I’ve never been a big fan of Sony’s A6000 series. But now that the A6700 has been out a while, I was keen to see if it lived up to some of the hype and how it compared to rival Canon and Fujifilm models. As you’ll see, it’s mostly good news with just a little bit of bad.

Body and handling

One of the things I’ve disliked about Sony’s APS-C bodies over the years is the usability and looks, especially compared to Fujifilm’s good-looking and easy-to-use models. I wouldn’t call the A6700 beautiful, but at least Sony has rectified the handling part.

The redesigned grip is larger and more comfortable, making it more comfortable to use over a full day. At the same time, Sony added a new control dial on the front, making the camera easier to use in manual or priority modes.

It includes a new dedicated photo, video and S&Q dial, letting you keep settings separate for each. Menus are a big step up too, as the A6700 uses the improved system from recent full-frame models. The only thing missing is a joystick, but the focus point can be adjusted using the d-pad-like dial on the back.

Steve Dent for Engadget

The A6700 is also the first Sony APS-C camera with an articulating display, so it’s better for vloggers than the flip-up display on past models. The relatively low resolution EVF is a weak point as it’s difficult at times to check focus, but it does the job most of the time. .

Another negative is the single card slot, but at least it supports high-speed UHS-II cards. Luckily, it has the same large battery as full-frame models, which gives it an excellent 570 shot CIPA rating.

Other features include microphone and headphone ports (along with support for Sony’s hot shoe audio accessories), a USB-C port for charging and data transfers and an HDMI port. The latter, unfortunately, is of the fiddly and fragile micro variety.

All of that adds up to a 6000-series camera I’d happily use for most types of work. Previously, I found those models not up to the job, especially for video.

Performance

Steve Dent for Engadget

Performance is more of a mixed bag, though. Lossless RAW bursts are possible at up to 11 fps, either in mechanical or electronic shutter modes. That compares to the 15 and 30fps for the similarly priced Canon EOS R7 and 15/20 fps for the Fujifilm X-T5. That’s quite a deficit considering the latter two have much higher resolution sensors.

The A6700 only stores up to 36 compressed RAW frames before the buffer fills, compared to 45 on the A6600 and comparable to rivals. Based strictly on speed, though, the R7 and X-T5 are better action cameras.

Fortunately, the autofocus is superb and that’s arguably more important for a consumer camera. In continuous mode, you’ll get reliable results even with fast moving subjects. And the AI tracking locks onto eyes and faces, ensuring you won’t miss important shots of rowdy kids, soccer games and more.

Steve Dent for Engadget

It also works with airplanes, animals, birds, cars or trains and insects. Unlike Canon’s auto system, though, you have to tell the A6700 exactly what you’re tracking. Once you’ve set it up the way you want, though, it’s a touch more reliable than Canon’s system, and significantly better than the X-T5..

The five-axis in-body stabilization is good but not great, offering 5 stops compared to 8 on the EOS R7 and 7 on the Fuji X-T5. Still, I was able to take sharp photos down to about an eighth of a second.

Rolling shutter was my main complaint with the A6600, but it’s now much improved and about as good as you get without a stacked sensor. It’s still present, though, so you’ll want to use the mechanical shutter for fast-moving subjects like propellers and trains.

Image Quality

With a new 26-megapixel sensor, The A6700 captures more detail than past 24-megapixel models. As mentioned, though, it’s lacking compared to the 32.5-megapixel Canon R7 and 40-megapixel X-T5.

Colors are mostly spot-on, but I still prefer Canon’s skin tones. JPEGs look good out of the camera, if a touch over-sharpened. With 14-bit RAW uncompressed files, I found plenty of room to adjust and tweak images, dialing down bright areas or adding detail to shadows. Keep in mind that when shooting bursts, though, RAW files are captured with only 12-bits of color fidelity.

There is a benefit to the lower resolution. The A6700 is better in low light than rivals, with noise well controlled up to ISO 6400. Images are usable up to ISO 12800, but anything beyond up to the ISO 102,400 limit is for emergency use only. It’s best to emphasize shadow exposure at high ISOs, as lifting those even a couple of stops creates excessive noise.

Video

The A6700 is so far above its predecessor for video that it’s useless to compare them. Rather, think of it as a cheaper, smaller FX30 cinema camera with the same image quality. The main difference is that the A6700 doesn’t have the same cooling capability, as I’ll discuss shortly.

It subsamples the full sensor width for 4K at 30 and 60 fps, so video is sharper than the X-T5 or R7. And the A6700 beats both of those models by having a 4K 120fps mode, though it’s cropped significantly at 1.58x.

Due to the relatively small body and lack of fans, you’ll need to be wary of overheating at 120p, as it’ll cut out after 20-30 minutes. For indoor shooting, you can generally shoot at up to 4K 60p without any stoppages with the “auto power temp” setting on “high.” Outside on a hot day, however, you may hit the limits and need to wait for the camera to cool down.

Like the FX30, you can shoot all video modes with 10-bit with S-Log3 capture. You can also load your own LUTs either to make log footage easier to monitor, or bake it into the final image.

Steve Dent for Engadget

Rolling shutter is still present, so you’ll still need to be careful with whip pans, fast subjects and the like. However, it’s far less bothersome than on past Sony crop sensor models.

It has a video feature that’s actually lacking on the FX3, namely auto-framing. That’s handy for vloggers as it can crop in and follow them as they move around the frame — with less quality loss than the ZV-E1 (this feature is finicky on the A6700 so be sure to test it first). It also offers focus compensation that digitally eliminates breathing, and again, this extra resolution compared to the ZV-E1 results in a sharper result.

Stabilization for video isn’t quite as good as the ZV-E1, though. The active mode is fine for handheld use and slow pans, but doesn’t do a lot to smooth out footsteps, and adds a 1.13x crop.

As for video quality, you’re seeing the same accurate colors and solid low-light capability as with photos. The 10-bit log options allow for plenty of flexibility in post, especially with contrasty images.

Wrap-up

Steve Dent for Engadget

The A6700 is easily Sony’s best APS-C camera yet, excelling at both photos and video, and offering much better handling. Despite being far more capable than the A6600, it carries the same $1,400 price.

As a photo camera, it’s slower than its main competition, the Canon R7 and Fujifilm X-T5, so those models are better for shooting action. Sony’s autofocus is slightly better though, and faster bursts are worthless if photos aren’t sharp.

As a video camera, though, it beats its main rivals across the board. All told, it’s a great option for content creators or hybrid shooters who favor video but do some photography. If that’s you, I’d highly recommend it.

Yesterday Google outlined the changes it will make to comply with the EU's Digital Markets Act (DMA) that goes into effect starting today. One important detail it left out, however, was whether it would charge developers who directed users outside the Play Store to sideload apps — and if so, how much. 

Now, Google has revealed that it will indeed charge developers even if they don't use the Play Store, just like Apple did with the App Store. Per new details found in the Play Console help section, the company will charge two new fees: 

1. An initial acquisition fee of 10% for in-app purchases or 5% for subscriptions for two years. This represents the value Play provided in facilitating initial user acquisition.

2. An ongoing services fee of 17% for in-app purchases or 7% for subscriptions. This covers ongoing Play services like parental controls, security, fraud prevention, and app updates.

Developers can opt out of ongoing fees after two years if users agree, but ongoing Play services will no longer apply. "Since users acquired the app through Play with the expectation of services such as parental controls, security scanning, fraud prevention, and continuous app updates, discontinuation of services requires user consent as well," Google stated.

Google included the following chart to show how the fees will apply to a hypothetical "Fantastiq App": 

Google

With this, Google is taking a similar approach to Apple, which reduced App Store commissions but introduced new fees. Namely, Apple tacked on on a new 3 percent “payment processing” fee for transactions that go through its store. And a new “core technology fee” will charge a flat €0.50 fee for all app downloads, regardless of whether they come from the App Store or a third-party website, after the first 1 million installations.

Google is justifying the fees by touting the value it provides in the Android ecosystem: "Play's fees support our investment in Android and Google Play and reflect the value provided by Android and Play, including enabling us to distribute Android for free and provide the continuously growing suite of tools and services that help developers build successful businesses, all while keeping our platforms safe and secure for billions of users worldwide."

Epic CEO Tim Sweeney already blasted Google's post about DMA compliance yesterday, before the new fees were even made public. "Google announced its malicious compliance plans for the European DMA law... it looks like their illegal anti-steering policy will be replaced by a new Google Tax on web transactions. We'll likely soon learn how he and other developers react to the new fees. 

Colorware has painted a lot of items over the years going back to the iPhone 3G and beyond, but its latest product is particularly clever. The limited edition Apple Number 2 Pencil transforms Apple's 2nd-gen stylus into a facsimile of a standard HB #2 pencil, with the only thing missing being chew marks on the eraser.

"What sets it apart? It connects with your iPad in a snap, charges without a hitch, and delivers precision that feels just right," the company states on the product page. "And it’s styled after the familiar pencil we all know, upgraded for today’s tech. It’s not just smart; it’s a smart looker too." (As a reminder, the 2nd-generation Apple Pencil offers features like "pixel-perfect precision, tilt and pressure sensitivity, and imperceptible lag," according to Apple.)

Colorware

Colorware goes on to describe other ways that it will change your life. "It’s that unspoken impact, the quiet ripple of intrigue it sends through the room. This isn’t just about having a piece of technology; it’s about carrying a symbol of uniqueness that speaks volumes before you do." 

That paint sure is doing a lot of work! But it ain't cheap — the Apple Number 2 Pencil costs $215, which is a $136 premium over the 2nd-generation Apple Pencil, currently priced at $79. Still, it's undeniably adorable and could be a great gift for an iPad artist. Just don't try to sharpen it (or chew on it). 

Last year TikTok rolled out a new monetization system for streamers called the Creativity Program to encourage longer videos that allow it to sell more ads. Now, the company is rolling the scheme out widely with a new name, the Creator Rewards Program, that only pays for videos longer than one minute. 

"The Creator Rewards Program will continue rewarding high-quality, original content over a minute long with an optimized rewards formula focused on 4 key areas: originality, play duration, search value and audience engagement," the company wrote. 

TikTok noted that longer content is more lucrative with “with total creator revenue increasing by over 250 percent within the last 6 months, and the number of creators making $50,000 each month nearly doubling” since the beta began. 

TikTok

TikTok is also expanding its subscription features for creators. Previously, only live streamers could access offerings like exclusive (paid) content, badges and personalized emoji, but now the company is expanding these benefits beyond live streams.

"In the coming weeks, eligible creators can sign up to access a new way to strengthen their community with added value through exclusive content and benefits, while providing their most engaged communities an opportunity to connect even deeper with their favorite creators," TikTok wrote.

The company’s Creator Fund, which had no minimum requirement for video length and ended last year, was often criticized for low payouts. Last year, streamer Hank Green shared that he made about 2.5 cents per 1,000 views on the platform — a fraction of his YouTube earnings and about half of what he earned on TikTok prior to the fund.

By comparison, select streamers embraced the beta Creativity Program. Some (with subscriber numbers varying from a half million to several million) received payouts ranging in the low thousands to nearly $100,000 per month, "a complete 180" from what they saw in the Creator Fund, according to one creator.

That said, audiences have been uncertain about longer videos. In a TikTok internal survey from last year, nearly 50 percent of users said videos over a minute in length were "stressful," and a third of users watched videos online at double speed, according to a Wired report from earlier this year.

How to pay creators is not TikTok's only challenge at the moment. Yesterday, a group of US lawmakers introduced a new bill that would force parent ByteDance to sell TikTok in order for the app to remain in the United States.

The battle between AI chatbots is more than a two-horse race. Anthropic, the company formed by several ex-OpenAI employees, claims its new Claude 3 language model outperforms ChatGPT and Google's Gemini in several key industry benchmarks. It even hit "near-human" levels on some tasks, the company wrote in a blog. 

There are three new chatbots under the Claude 3 umbrella, including Haiku, Sonnet, and Opus. Sonnet powers the Claude.ai chatbot and is offered for free with an email sign-in. Meanwhile, Opus is the largest and most powerful LLM and will be available with a $20 per month subscription via the "Claude Pro" service. It's also multi-modal, so it can work with both text and image inputs, unlike past versions.

All Claude 3 models "can power live customer chats, auto-completions and data extraction tasks where responses must be immediate and in real-time," the company said. On top of promising "near-instant results," they can supposedly handle longer, multi-step instructions with increased accuracy.

Anthropic

Opus showed better graduate-level reasoning than GPT-4, scoring 14.7 percent higher in that test than GPT-4. It also beat OpenAI's chatbot in tasks involving math, coding, reasoning and knowledge. 

They also top past Claude models. "For the vast majority of workloads, Sonnet is 2x faster than Claude 2 and Claude 2.1 with higher levels of intelligence. It excels at tasks demanding rapid responses, like knowledge retrieval or sales automation. Opus delivers similar speeds to Claude 2 and 2.1, but with much higher levels of intelligence," according to Anthropic.

Meanwhile Haiku, the smallest version of Claude 3, is "the fastest and most cost-effective model on the market." To that end, it's capable of reading a dense research paper complete with charts and graphs in under three seconds. 

The company also noted that Claude 3 "can process a wide range of visual formats, including photos, charts, graphs and technical diagrams," aiding companies that use PDFs, flowcharts, or presentation slides. It'll also be less likely to refuse harmless content thanks to a more nuanced understanding of requests, while still recognizing "real harm."

Anthropic has said that Claude AI is guided by 10 secret foundational pillars of fairness. Claude 3 was trained on both nonpublic internal and public-facing data, using hardware from Amazon Web Services (AWS) and Google Cloud (Amazon recently invested $4 billion in Anthropic). 

Claude 3 Opus and Claude 3 Sonnet are available now through Anthropic's API, with Haiku set to follow soon. Sonnet is also accessible through Amazon Bedrock and in private preview on Google Cloud's Vertex AI Model Garden.

Ford EVs will gain access to Tesla's Superchargers across the US and Canada starting today, becoming the first non-Tesla vehicles to do so, Ford announced. The companies cemented the fast charging pact last year and numerous other automakers followed suit later on. The expanded network will be a big benefit for the Ford and its customers amid news that sales of its EVs have heavily sagged.

Current Ford EVs use CCS type chargers, so current customers must order the Fast Charging Adapter (NACS) compatible with Tesla's Superchargers (below). That will be available at no charge to new and existing clients in the BlueOval charge network until June 30, 2024. After that, it will cost $230 including estimated taxes and shipping. Starting in 2025, Ford EVs will come standard with the NACS charging system. 

Ford

Customers can use the FordPass App to locate the new Tesla Superchargers, just as they would when looking for BlueOval chargers. Tesla Superchargers will also be coming to Apple Maps EV routing, Google Maps EV routing, and the Ford Connected built-in navigation BlueOval charge network.

As for payment, you'll be able to handle that through the FordPass app and Charge Assist app in the vehicle's touchscreen. "This means customers simply have to plug in and charging will automatically start with [costs] managed through FordPass." BlueOval charge network membership is required, and if you're not yet enrolled, you'll be prompted to do so when order the NACS adapter (check's Ford's website for more details).

Tesla's 15,000+ strong Supercharger network will more than double Ford EV owners' access to fast DC chargers. With that, it will have 28,000 fast chargers and 126,000 chargers total. The company notes that a lack of charging stations is the second largest barrier after price for customers potentially selecting EVs instead of ICE or PHEV vehicles. 

The news is good for Ford and other automakers who will get a big boost in the number of fast charging spots. However, Tesla owners will see a lot more Supercharger competition. Last year, the US government opened $2.5 billion in funding for community EV chargers

With declining EV sales, Ford was recently forced to lower Mach E prices by $3,100 to $8,100, it recently said in a financial statement. The company also cut production of that model and the F-150 Lightning pickup truck due to the slowdown.