Author Archives: Jon Fingas

Meta’s Oversight Board recommends stronger rules on gender-based violence

Posted on by

Meta's Oversight Board is calling for changes to the social media giant's policies on gender-based violence. In addition to overturning the decision to preserve a Facebook post making light of domestic abuse, the panel has recommended Meta create policies cracking down on content that "normalizes" the violence by celebrating, justifying or mocking it. The board members also want Meta to clarify that its anti-harassment provisions include serious physical injuries as "medical condition[s]."

The findings come in response to a 2021 Iraq post 'joking' about a woman being beaten for making a mistake writing to her husband. The woman in the photo was a Syrian activist, and the post contained a hashtag used for pro-women discussions in the country. A Facebook user reported the content three times in February this year, but the report was automatically closed as the complaint wasn't reviewed. Meta only pulled the post after the Oversight Board selected the user's appeal for review.

The board is concerned the post wouldn't have violated Meta's policies on harassment if the woman were fictional or couldn't be identified. Moreover, the post went untouched for two years and didn't get a human review when it was originally reported earlier this year. This implies Meta "does not prioritize" this kind of violation, according to the board.

We've asked Meta for comment on the ruling, including whether or not it intends to implement the recommendations. The company has adopted Oversight Board recommendations before, although it has been criticized for an alleged lack of transparency and delays in sharing information.

The Oversight Board has requested significant changes to Meta's behavior in recent months. It asked Meta to rethink the COVID-19 misinformation it removes, and has pressed for more inclusive rules on adult nudity. In June, the panel even demanded Facebook suspend Cambodia's Prime Minister for threatening violence against political opponents.

This article originally appeared on Engadget at https://www.engadget.com/metas-oversight-board-recommends-stronger-rules-on-gender-based-violence-152712977.html?src=rss

YouTube tests AI-generated video summaries

Posted on by

The next internet video summary you see might not have been written by a human. YouTube is testing AI-generated video summaries that appear on search and watch pages. The text is meant to provide a "quick overview" of a clip to help you decide if it's worth watching. The company is quick to stress that these don't replace creators' own video descriptions.

The test will only be visible with a "limited number" of English-language videos, and only for some viewers. In a statement to Engadget, a YouTube spokesperson says the summaries will appear on mobile devices for how-to videos, shopping videos and vlogs. You'll see the videos worldwide. 

YouTube has given users a few preview features in recent weeks, as Android Policeexplains. Premium subscribers could lock the screen during playback to prevent accidental commands. You've also had the option to turn Shorts comments into entirely new short-form clips.

The experiment comes nearly three years after YouTube started testing AI-produced video chapters, but it's part of a larger generative AI push at Google. The tech giant has introduced its Bard chatbot and is using the technology to produce everything from spreadsheet templates to whole news articles, if not always for public consumption.

The YouTube summary feature may be one of the more logical extensions, however. The service says over 500 hours of content are uploaded every minute — it would be impossible for humans to keep up. The question is whether or not the AI summaries are accurate enough. Google has warned that generative systems like Bard may be prone to inaccuracies and misinformation, and it's not yet clear how well the YouTube experiment works in practice.

Update 8/1 11:48AM ET: We've updated the story with a statement from YouTube on where you'll see the summaries, including countries, platforms and video types.

This article originally appeared on Engadget at https://www.engadget.com/youtube-tests-ai-generated-video-summaries-140259921.html?src=rss

Uber safety driver involved in fatal self-driving car crash pleads guilty

Posted on by

The Uber safety driver at the wheel during the first known fatal self-driving car crash involving a pedestrian has pleaded guilty to and been sentenced for an endangerment charge. Rafaela Vasquez will serve three years of probation for her role in the 2018 Tempe, Arizona collision that killed Elaine Herzberg while she was jaywalking at night. The sentence honors the prosecutors' demands and is stiffer than the six months the defense team requested.

The prosecution maintained that Vasquez was ultimately responsible. While an autonomous car was involved, Vasquez was supposed to concentrate on the road and take over if necessary. The modified Volvo XC90 in the crash was operating at Level 3 autonomy and could be hands-free in limited conditions, but required the driver to take over at a moment's notice. It noticed Herzberg but didn't respond to her presence.

The defense case hinged on partly blaming Uber. Executives at the company thought it was just a matter of time before a crash occurred, according to supposedly leaked conversations. The National Transportation Safety Board's (NTSB) collision findings also noted that Uber had disabled the emergency braking system on the XC90, so the vehicle couldn't come to an abrupt stop.

Tempe police maintained that Vasquez had been watching a show on Hulu and wasn't paying attention during the crash. Defense attorneys have insisted that Vasquez was paying attention and had only been momentarily distracted.

The plea and sentencing could influence how other courts handle similar cases. There's long been a question of liability surrounding mostly driverless cars — is the human responsible for a crash, or is the manufacturer at fault? This suggests humans will still face penalties if they can take control, even if the punishment isn't as stiff for conventional situations.

Fatal crashes with autonomy involved aren't new. Tesla has been at least partly blamed for collisions while Full Self Driving was active. The pedestrian case is unique, though, and looms in the background of more recent Level 4 (fully driverless in limited situations) offerings and tests from Waymo and GM's Cruise.While the technology has evolved since 2018, there are still calls to freeze robotaxi rollouts over fears the machines could pose safety risks.

This article originally appeared on Engadget at https://www.engadget.com/uber-safety-driver-involved-in-fatal-self-driving-car-crash-pleads-guilty-212616187.html?src=rss

Mazda stops selling its only EV in the US

Posted on by

Mazda's MX-30 is a strange EV, and it appears the company knows this. The automaker is pulling the small crossover from the US market after the 2023 model year. The company says it's currently focused on "large platform" plug-in hybrids like the CX-70 and CX-90, and also offers conventional hybrids like the CX-50.

The MX-30 is Mazda's first and only mass-produced EV to date, and it came to the country in fall 2021 as a California-only model. From the start, its prospects were limited by a number of unusual business decisions. On top of its small size, it had just a 35.5kWh battery with an estimated 100-mile range. This was supposed to make it feel more like a gas car and limit the car's CO2 footprint, but the net result was a car that was both less exciting to drive and less practical than competing EVs costing significantly less than Mazda's $35,385 sticker. The outgoing current-gen Chevy Bolt costs $9,000 less while delivering more performance and over twice the range.

Sales were accordingly poor. Mazda has sold just 571 MX-30 units in the US over three years. In some countries, the company has reworked the machine as a plug-in hybrid with a rotary gas engine and an estimated 53 miles of electric-only driving.

Mazda is rethinking its EV strategy. It plans to introduce cars based on both an existing platform as well as a new framework. There are claims the company might use a future Toyota platform expected to deliver over 600 miles on a charge. The MX-30 discontinuation isn't the end, in other words — rather, it's an acknowledgment that the company needs to be more competitive.

The move is also a reflection of Japanese brands' general trouble entering the EV market. Honda might not replace its E compact car, and is refocusing on SUVs like next year's Prologue as well as its collaboration with Sony. Toyota and Subaru, meanwhile, had problems launching the bZ4x SUV crossover and its Solterra equivalent. For now, they largely have to rely on hybrids to attract buyers.

This article originally appeared on Engadget at https://www.engadget.com/mazda-stops-selling-its-only-ev-in-the-us-182657731.html?src=rss

‘Minecraft’ mod exploit lets hackers control your device

Posted on by

You might want to run antivirus tools if you use certain Minecraft mods. The MMPA security community has learned that hackers are exploiting a "BleedingPipe" flaw in the Forge framework powering numerous mods, including some versions of Astral Sorcery, EnderCore and Gadomancy. If one of the game tweaks is running on Forge 1.7.10/1.12.2, intruders can remotely control both servers and gamers' devices. In one case, an attacker was using a new exploit variant to breach a Minecraft server and steal both Discord chatters' credentials as well as players' Steam session cookies.

As Bleeping Computerexplains, BleedingPipe relies on incorrect deserialization for a class in the Java code powering the mods. Users just have to send special network traffic to a server to take control. The first evidence of BleedingPipe attacks surfaced in March 2022 and were quickly patched by modders, but MMPA understands most servers running the mods haven't updated.

We've asked Mojang parent company Microsoft for comment. It's not responsible for Forge, so the tech giant can't necessarily stop or limit the damage. You won't be affected if you use stock Minecraft or stick to single-player sessions.

The full scope of the vulnerability isn't clear. While there are 46 mods known to fall prey to BleedingPipe as of this writing, there's the potential for considerably more. Users are asked to scan their systems (including their Minecraft folder) for malware. Server operators, meanwhile, are urged to either update mods or stop running them entirely. MMPA also has a PipeBlocker mod that protects everyone involved, although mod packs may cause problems if the mods haven't been updated.

This article originally appeared on Engadget at https://www.engadget.com/minecraft-mod-exploit-lets-hackers-control-your-device-162231445.html?src=rss

Twitter threatens to sue anti-hate group over its research

Posted on by

X Corp (aka Twitter) may take legal action against one of its more outspoken critics. The Center for Countering Digital Hate (CCDH) says in a letter that X threatened a lawsuit against the anti-hate group on July 20th for allegedly making "false or misleading" claims against the social media giant and trying to scare advertisers. The Center published a research article in June asserting that X allowed explicitly racist and homophobic posts despite policies to the contrary, even days after they'd been reported.

X accused CCDH of using poor methodology, and not studying the 500 million posts on the service each day. It also maintained that the Center was taking funding from competitors or foreign governments as part of an "ulterior agenda," according to The New York Times. CCDH rejects the allegations. It notes that it never claimed to be conducting a comprehensive study, and points to its documented methodology. It adds that X never said just what was inaccurate, and that it doesn't accept any funding from companies or governments.

The researchers further maintain that X is being hypocritical for attacking supposedly limited research while simultaneously curbing the ability to conduct those studies. The platform recently instituted reading rate limits in what it says is a temporary measure to thwart excessive data scraping. The company believes some scrapers are abusing their tools to train AI models and conduct manipulation campaigns. Even Blue subscribers are capped at viewing 8,000 posts per day, making it impractical to conduct extensive research.

X has disbanded its communications team and isn't available for comment. CCDH says it "will not be bullied" and will continue publishing its research. It also intends to post the original letter, and believes a lawsuit with "frivolous" claims could prove risky.

Reports indicate that X's ad sales have plunged by half since Musk bought the company last year. The executive pinned the exodus on European and North American marketers trying to deliberately bankrupt the firm, but employees talking to the NYT maintained that advertisers were balking at the surges in hate speech and porn after Musk's acquisition. Brands like GM and Volkswagen have frozen ad spending on X, while others are believed to have scaled back their efforts.

X has been threatening legal action against others in recent weeks. It accused Microsoft of violating data use policy, and threatened to sue Meta for purportedly duping key features with Threads. The tech giant has also sued a law firm for supposedly taking excess funds during the previous management's handover to Musk.

This article originally appeared on Engadget at https://www.engadget.com/twitter-threatens-to-sue-anti-hate-group-over-its-research-125645342.html?src=rss

‘Final Fantasy XIV’ comes to Xbox next spring

Posted on by

It took a decade, but Square Enix's premier massively multiplayer online role-playing game is finally coming to Xbox consoles. The developer has revealed that Final Fantasy XIV will be available for Xbox Series X/S in spring 2024. Like its PS5 counterpart, this version will support 4K visuals on Series X and faster loading times. It's not yet clear if there will be Xbox-only upgrades.

An open beta is expected for patch 6.5X. In other words, the Xbox port should be ready in time for the Dawntrail expansion due in summer next year.

Microsoft has been eager to add Final Fantasy games to its catalog. In 2019, it added 10 titles to Game Pass that included many of the releases from VII through to XV. The deluge didn't include XIV, however, leaving Xbox players without an active MMO. The game debuted on PS3 and PC in 2013, with ports for PS4 (2014), Mac (2015) and PS5 (2021) in subsequent years.

The incentives are clear. Final Fantasy XIV helps court fans of the series, particularly those left out by the timed PS5 exclusive for XVI. It's also an attempt to reach out to both Japanese gamers and JRPG enthusiasts. The Xbox has struggled in Japan due in no small part to local studios skipping the platform in favor of domestic consoles from Nintendo and Sony. This game won't suddenly improve Microsoft's fortunes, but it does eliminate a barrier to adoption for some players.

This article originally appeared on Engadget at https://www.engadget.com/final-fantasy-xiv-comes-to-xbox-next-spring-192903645.html?src=rss

iRobot’s Roomba s9+ robot vacuum is back down to a record low price

Posted on by

If you've wanted a top-tier robot vacuum but were put off by the sometimes outlandish prices, now's your time to act. Amazon is once more selling iRobot's Roomba s9+ at a record low of $600, or 40 percent off. That's the best we've seen since Prime Day, and puts it at the same price as a mid-range model like the Roomba j7 or Shark's AI Ultra.

The Roomba s9+ is still our pick for the best premium robot vacuum. It's powerful, good at navigating floors (both carpet and hardwood), tackles corners well and empties itself relatively quietly. The slick design also makes it look right at home in an upscale abode. You shouldn't have to worry about it getting stuck or missing an important mess.

It's not flawless. The Roomba s9+ isn't specifically geared to avoid pet poop, so you may need to look elsewhere if your cat or dog routinely leaves unwanted surprises. It's also louder than newer (if less effective) options like the j7. At this price, though, it's an easy choice — you'll get efficient cleaning that frees you for more important tasks.

Follow @EngadgetDeals on Twitter and subscribe to the Engadget Deals newsletter for the latest tech deals and buying advice.

This article originally appeared on Engadget at https://www.engadget.com/irobots-roomba-s9-robot-vacuum-is-back-down-to-a-record-low-price-152658325.html?src=rss

WhatsApp adds Telegram-style video messages

Posted on by

It's now easy to show your WhatsApp friends what you're doing without putting in the effort to create a Story or a similar clip. The service is rolling out Telegram-like instant video messages that let you respond in chat with videos up to a minute long. You just have to tap and hold a button to capture footage, and you can swipe up to keep the recording going while hands-off. 

Videos will automatically play on mute, so you shouldn't startle anyone until you're ready to turn on sound with a tap. Like most anything else in WhatApp, the instant messages are end-to-end encrypted to add a layer of privacy. The feature should be available to all users in the "coming weeks," the company says.

This likely won't replace Telegram Snapchat for many people. WhatsApp is still dominant in the messaging space, and driven largely by calls and texts. Even so, the feature might be appreciated if you'd rather not switch apps just to send rapid-fire videos. You can quickly share news when text or an audio message won't cut it, or simply react to a friend with something more personal than a sticker.

There's an incentive for parent company Meta to act. While WhatsApp has over 2 billion active users as of last year, Telegram is expanding quickly with an estimated 700 million active users each month. Snapchat is still growing despite fierce competition from other social platforms. Apple's iOS 17 will introduce FaceTime video voicemail to iPhone users, too. While instant video messages might not persuade users to sign up, it might keep WhatsApp fans from drifting toward the competition.

This article originally appeared on Engadget at https://www.engadget.com/whatsapp-adds-telegram-style-video-messages-141005605.html?src=rss

Android’s earthquake warning system failed in Turkey, according to the BBC

Posted on by

Google's earthquake warning system for Android is supposed to provide notices in time to reach safety, but that might not have happened following the quake in Turkey on February 6th. BBC investigators claim that none of the hundreds of people they talked to in three Turkish cities received an alert before the first tremor hit. Only a "limited number" got an alert for a second tremor, investigators say.

In a statement to Engadget, Google says "millions" of people in Turkey received earthquake alerts, although the company hasn't shared data indicating widespread notifications. Google did show the BBC a handful of social media posts from people who said they received a warning, but only one was for the first quake. Product lead Micah Berman tells the outlet he doesn't have a "resounding answer" as to why social networks were quiet about alerts. A spokesperson tells Engadget the technology is "supplemental," though, and not meant to replace conventional warning systems.

The Android Earthquake Alert System uses the accelerometer (that is, motion sensing) in phones to effectively crowdsource warnings. If many phones vibrate at the same time, Google can use the collective data to find the epicenter and magnitude of the quake, automatically sending a warning to people who are likely to feel the brunt of the shaking. While there's no more than a minute's notice, that can be enough time to find cover or evacuate. The technology can theoretically help people in areas where normal warnings are unavailable.

The concern is that the system might have failed during a strong (7.8-magnitude) earthquake. Even if it worked, it's not clear how many people should and do receive warnings in cases like this, not to mention milder incidents. Without more data, it's not certain that Android's quake alerts will consistently reach enough people to have an impact.

Update 7/28 10:24AM ET: Google has issued a statement to Engadget with more details, including an emphasis that the alerts are supplemental. We've updated the story accordingly.

This article originally appeared on Engadget at https://www.engadget.com/androids-earthquake-warning-system-failed-in-turkey-according-to-the-bbc-210612253.html?src=rss