verge-rss

Image: Bethesda

If you haven’t played Doom or Doom II for a while — or ever — a new re-release that Bethesda surprise-dropped (sorta) on Thursday might be the perfect excuse to jump in to the classic games. The re-release, which combines both games into one package called Doom + Doom II and is a free update for anyone who already owns Doom (1993) or Doom II, offers a long list of great new features — including a brand new single-player episode and online, cross-platform deathmatch multiplayer.
With Doom + Doom II, you’ll have access to both of those two games as well as extra single-player content like John Romero’s Sigil episode released in 2019 and Legacy of Rust, which is a new Doom episode created by “individuals from id Software, Nightdive Studios and MachineGames,” according to Bethesda. You’ll also be able to download single-player mods through a new in-game mod browser.

Image: Bethesda

On the multiplayer front, Bethesda has added 25 new maps for deathmatch, meaning you’ll have up to 43 to pick from. The cross-platform deathmatch and co-op also support up to 16 players.
The combined package will support up to 4K resolution and 120fps if you’re playing on PC, PlayStation, or Xbox Series X / S. There’s also updated music for the games by Andrew Hulshult (who worked on the music for Doom Eternal: The Ancient Gods DLC), and you can switch back and forth between the newer tracks and older tracks whenever you like.
There’s a lot more I didn’t touch on, and if you’re at all interested in this package, I recommend scrolling through the extensive patch notes. The game is available on PC (via Steam, the Windows Store, the Epic Games Store, and GOG), Xbox One, Xbox Series X / S, PS4, PS5, and Nintendo Switch.
There was one other cool piece of retro Doom news announced today, too: Limited Run Games is working on a re-release of Doom on the SNES. The game is set to release in 2025 — and, naturally, there will only be 666 copies made.

The new. The improved. The DEFINITIVE DOOM SNES experience! All 4 DOOM episodes, including Thy Flesh Consumed Added circle-strafe Performance improvements Monster respawning on Nightmare difficulty Level codes Rumble support via an all-new controller!#QuakeCon pic.twitter.com/ylaYZ8JbMg— Limited Run Games (@LimitedRunGames) August 8, 2024

Image: Bethesda

If you haven’t played Doom or Doom II for a while — or ever — a new re-release that Bethesda surprise-dropped (sorta) on Thursday might be the perfect excuse to jump in to the classic games. The re-release, which combines both games into one package called Doom + Doom II and is a free update for anyone who already owns Doom (1993) or Doom II, offers a long list of great new features — including a brand new single-player episode and online, cross-platform deathmatch multiplayer.

With Doom + Doom II, you’ll have access to both of those two games as well as extra single-player content like John Romero’s Sigil episode released in 2019 and Legacy of Rust, which is a new Doom episode created by “individuals from id Software, Nightdive Studios and MachineGames,” according to Bethesda. You’ll also be able to download single-player mods through a new in-game mod browser.

Image: Bethesda

On the multiplayer front, Bethesda has added 25 new maps for deathmatch, meaning you’ll have up to 43 to pick from. The cross-platform deathmatch and co-op also support up to 16 players.

The combined package will support up to 4K resolution and 120fps if you’re playing on PC, PlayStation, or Xbox Series X / S. There’s also updated music for the games by Andrew Hulshult (who worked on the music for Doom Eternal: The Ancient Gods DLC), and you can switch back and forth between the newer tracks and older tracks whenever you like.

There’s a lot more I didn’t touch on, and if you’re at all interested in this package, I recommend scrolling through the extensive patch notes. The game is available on PC (via Steam, the Windows Store, the Epic Games Store, and GOG), Xbox One, Xbox Series X / S, PS4, PS5, and Nintendo Switch.

There was one other cool piece of retro Doom news announced today, too: Limited Run Games is working on a re-release of Doom on the SNES. The game is set to release in 2025 — and, naturally, there will only be 666 copies made.

Read More 

Illustration: The Verge

The Turkish government has blocked access to Roblox following a government investigation that found content that could lead to child exploitation. “According to our Constitution, our State is obliged to take the necessary measures to ensure the protection of our children,” Justice Minister Yılmaz Tunç says in a machine-translated post on X.
It’s unclear when Roblox might be unblocked in the country. “On Aug 7, we learned that Roblox was blocked in Türkiye, and we are working urgently to understand the issue with the goal of resolving it,” the company said in a statement on its website. “We are mindful that millions of people use our platform in Turkey everyday, including a vibrant community of developers whose businesses rely on Roblox. We’re always thinking about how to keep our community safe, and we share global policymakers’ commitment to protecting children.
“We respect the laws and regulations in countries where we operate and share local lawmakers’ commitment to children,” Roblox spokesperson Eric Porterfield says in a statement to The Verge. “We look forward to working together to ensure Roblox is back online in Türkiye as soon as possible.”
The Roblox block follows Turkey’s decision to shut down access to Instagram last week. Turkey’s Information and Communication Technologies Authority didn’t provide a reason for that block at the time, according to The Associated Press. In April, Meta announced that it would temporarily shut down access to Threads in Turkey due to a regulatory decision. The country also overturned a ban on Wikipedia after more than two years in 2020.

Illustration: The Verge

The Turkish government has blocked access to Roblox following a government investigation that found content that could lead to child exploitation. “According to our Constitution, our State is obliged to take the necessary measures to ensure the protection of our children,” Justice Minister Yılmaz Tunç says in a machine-translated post on X.

It’s unclear when Roblox might be unblocked in the country. “On Aug 7, we learned that Roblox was blocked in Türkiye, and we are working urgently to understand the issue with the goal of resolving it,” the company said in a statement on its website. “We are mindful that millions of people use our platform in Turkey everyday, including a vibrant community of developers whose businesses rely on Roblox. We’re always thinking about how to keep our community safe, and we share global policymakers’ commitment to protecting children.

“We respect the laws and regulations in countries where we operate and share local lawmakers’ commitment to children,” Roblox spokesperson Eric Porterfield says in a statement to The Verge. “We look forward to working together to ensure Roblox is back online in Türkiye as soon as possible.”

The Roblox block follows Turkey’s decision to shut down access to Instagram last week. Turkey’s Information and Communication Technologies Authority didn’t provide a reason for that block at the time, according to The Associated Press. In April, Meta announced that it would temporarily shut down access to Threads in Turkey due to a regulatory decision. The country also overturned a ban on Wikipedia after more than two years in 2020.

Read More 

Illustration by The Verge

Anybody with a Substack account can now publish posts on Substack’s web and mobile app, the company announced on Thursday, turning the newsletter platform into even more of a social network. (Currently, users publishing posts on their profile from mobile can only do so via the Substack iOS app, though support for Android is coming soon.)
Substack members can publish content, get paid and free subscribers, and accumulate pledges from their Substack profile without needing to create a publication. Creating a publication, however, allows them to take advantage of more advanced features like a website, multiple admins, and sections. If they do choose to start a new one, Substack says, they can import any subscribers they may have accumulated.
Substack has been increasingly turning the publishing platform into a social network
Users can’t make their Substack profiles private, however, so anybody can follow and read posts. If they want to publish posts for a limited audience, they’d have to create and link to a private publication on their profile.
Along with these updates, Substack also announced that it’s continuing to work on making the platform more mobile-friendly. Last month, the platform started letting users publish full posts from the mobile app. Today, the company revealed it’s also testing live video and in-app payments.
Last year, the company arguably turned into an X rival when it launched its Notes feature for publishing short posts.

Illustration by The Verge

Anybody with a Substack account can now publish posts on Substack’s web and mobile app, the company announced on Thursday, turning the newsletter platform into even more of a social network. (Currently, users publishing posts on their profile from mobile can only do so via the Substack iOS app, though support for Android is coming soon.)

Substack members can publish content, get paid and free subscribers, and accumulate pledges from their Substack profile without needing to create a publication. Creating a publication, however, allows them to take advantage of more advanced features like a website, multiple admins, and sections. If they do choose to start a new one, Substack says, they can import any subscribers they may have accumulated.

Users can’t make their Substack profiles private, however, so anybody can follow and read posts. If they want to publish posts for a limited audience, they’d have to create and link to a private publication on their profile.

Along with these updates, Substack also announced that it’s continuing to work on making the platform more mobile-friendly. Last month, the platform started letting users publish full posts from the mobile app. Today, the company revealed it’s also testing live video and in-app payments.

Last year, the company arguably turned into an X rival when it launched its Notes feature for publishing short posts.

Read More 

Image: The Verge

OpenAI is going to let users of ChatGPT’s free tier make up to two images per day with its DALL-E 3 model, the company announced on Thursday. When the company launched DALL-E 3 in September, it was available first to users who paid for ChatGPT Plus.
One of DALL-E 3’s key improvements is that ChatGPT can come up with a prompt to make an image, which should make it easier to make images. For example, as my former colleague Emilia David wrote last year:
In a demo to The Verge, Aditya Ramesh, lead researcher and head of the DALL-E team, prompted ChatGPT to help him come up with a logo for a ramen restaurant in the mountains. ChatGPT then wrote a longer prompt, and DALL-E came up with four options.

We’re rolling out the ability for ChatGPT Free users to create up to two images per day with DALL·E 3. Just ask ChatGPT to create an image for a slide deck, personalize a card for a friend, or show you what something looks like. pic.twitter.com/3csFTscA5I— OpenAI (@OpenAI) August 8, 2024

OpenAI says the ability to create images with DALL-E 3 is “rolling out,” but you might already have access: while writing this article, I was able to make two images with the ChatGPT Mac app before getting a notice that I had reached my image creation limit for the day.
It’s been a busy day for OpenAI news. The company released a safety assessment of its GPT-4o model, added a new person to its board of directors, and CEO Sam Altman was sent a letter from Democrats in Congress pushing for answers about OpenAI’s safety record.

Image: The Verge

OpenAI is going to let users of ChatGPT’s free tier make up to two images per day with its DALL-E 3 model, the company announced on Thursday. When the company launched DALL-E 3 in September, it was available first to users who paid for ChatGPT Plus.

One of DALL-E 3’s key improvements is that ChatGPT can come up with a prompt to make an image, which should make it easier to make images. For example, as my former colleague Emilia David wrote last year:

In a demo to The Verge, Aditya Ramesh, lead researcher and head of the DALL-E team, prompted ChatGPT to help him come up with a logo for a ramen restaurant in the mountains. ChatGPT then wrote a longer prompt, and DALL-E came up with four options.

OpenAI says the ability to create images with DALL-E 3 is “rolling out,” but you might already have access: while writing this article, I was able to make two images with the ChatGPT Mac app before getting a notice that I had reached my image creation limit for the day.

It’s been a busy day for OpenAI news. The company released a safety assessment of its GPT-4o model, added a new person to its board of directors, and CEO Sam Altman was sent a letter from Democrats in Congress pushing for answers about OpenAI’s safety record.

Read More 

Photo by Tom Warren / The Verge

It’s a mess, but Intel and partners are working to clean it up. Many months ago, gamers began experiencing strange crashes with their 13th and 14th Gen Intel Core i9 CPUs — but that was just the tip of the iceberg.
Intel has now extended its warranty by two full years on 24 different 13th Gen and 14th Gen desktop chips, including Core i5, Core i7, and Core i9 models, after discovering that many CPUs based on its Raptor Lake architecture are susceptible to permanent damage. They were being fed too much voltage, and some have irreversibly degraded. Intel has not yet said if laptop chips are failing the same way.
As of August 2024, there is currently no fix for an Intel CPU that’s crashing like this — you need to exchange it, and that’s what the extended two-year warranty is for. But you can apply motherboard BIOS updates that may prevent damage to begin with, and Intel is explicitly addressing the excessive voltage issue in August updates for its partners’ motherboards.
There are many open questions, like how many customers are affected, why Intel isn’t recalling these chips, why it didn’t claw back existing inventory ahead of its August patch, how it will proactively warn customers, how lenient it might be when dealing with customers who want an exchange, and which other issues might be contributing to these chips’ instability. (Intel has suggested the voltage is a primary cause, but not the only one.)
The Verge is keeping track of the answers, and other news, in updates to the StoryStream you’ll find below. We’re also keeping track of which PC makers will honor Intel’s extended warranty in this story.

Photo by Tom Warren / The Verge

It’s a mess, but Intel and partners are working to clean it up.

Many months ago, gamers began experiencing strange crashes with their 13th and 14th Gen Intel Core i9 CPUs — but that was just the tip of the iceberg.

Intel has now extended its warranty by two full years on 24 different 13th Gen and 14th Gen desktop chips, including Core i5, Core i7, and Core i9 models, after discovering that many CPUs based on its Raptor Lake architecture are susceptible to permanent damage. They were being fed too much voltage, and some have irreversibly degraded. Intel has not yet said if laptop chips are failing the same way.

As of August 2024, there is currently no fix for an Intel CPU that’s crashing like this — you need to exchange it, and that’s what the extended two-year warranty is for. But you can apply motherboard BIOS updates that may prevent damage to begin with, and Intel is explicitly addressing the excessive voltage issue in August updates for its partners’ motherboards.

There are many open questions, like how many customers are affected, why Intel isn’t recalling these chips, why it didn’t claw back existing inventory ahead of its August patch, how it will proactively warn customers, how lenient it might be when dealing with customers who want an exchange, and which other issues might be contributing to these chips’ instability. (Intel has suggested the voltage is a primary cause, but not the only one.)

The Verge is keeping track of the answers, and other news, in updates to the StoryStream you’ll find below. We’re also keeping track of which PC makers will honor Intel’s extended warranty in this story.

Read More 

The Thumby Color shrinks the Game Boy Advance to a handheld just two inches in size. | Image: TinyCircuits

How small can you shrink the Nintendo Game Boy Advance? That’s a question Tiny Circuits is answering with the Thumby Color, a GBA clone measuring just two inches in size that still manages to fit a directional pad, shoulder buttons, a rumble motor, a 16-bit color screen, and the ability to link two of them together with a USB-C cable for multiplayer gaming.
Unlike the myriad of Game Boy and GBA clones released by companies like Anbernic, the Thumby Color is not an emulator. It doesn’t have enough processing power to play retro games nor enough storage to hold hundreds of ROM files.
It’s powered by a 150MHz Raspberry Pi RP2350 processor that’s paired with just 16MB of storage — yes, that’s megabytes, not gigabytes. Not being able to play actual GBA games on the Thumby Color is probably a good thing, because its 0.85-inch square LCD screen has a resolution of just 128 x 128 pixels, which is less than half the resolution of the GBA.
Games are instead programmed in MicroPython. The Thumby Color will come with six open source titles pre-installed, including a first-person shooter called Monstra and a dungeon crawler called Thumbgeon 2. It will also come with all the monochromatic games that shipped with the original Thumby back in 2021, which was the company’s ultra-tiny clone of the original Game Boy.
The Thumby Color isn’t a handheld you’re going to want to play for hours on end, but it is one you can potentially sneak into any setting where you want a distraction — without trying to hide more conspicuous handhelds like the Steam Deck. It’s also a good excuse to learn a programming language as TinyCircuits has created tutorials and guides available for free through its website.

Image: TinyCircuits
The Thumby Color is available in several different colorways including transparent case options.

TinyCircuits has opted for a Kickstarter crowdfunding campaign to bring the Thumby Color to consumers. The original Thumby still sells for $42, but this new color screen version is available with a pledge of $49 for a purple version, with shipping expected as early as November. If you want a Thumby Color with a clear or transparent purple housing, the price jumps to $55, and if you want a Kickstarter exclusive Black version, it’s $75.
Keep in mind that there’s always the chance a crowdfunded product will not deliver all the features promised at the time of the campaign. It could also be significantly delayed, or never delivered at all, without refunds. TinyCircuits has successfully shipped several Kickstarted products before, so there’s slightly less risk here, but there’s no guarantee the product will materialize even with its funding goal already being met.

The Thumby Color shrinks the Game Boy Advance to a handheld just two inches in size. | Image: TinyCircuits

How small can you shrink the Nintendo Game Boy Advance? That’s a question Tiny Circuits is answering with the Thumby Color, a GBA clone measuring just two inches in size that still manages to fit a directional pad, shoulder buttons, a rumble motor, a 16-bit color screen, and the ability to link two of them together with a USB-C cable for multiplayer gaming.

Unlike the myriad of Game Boy and GBA clones released by companies like Anbernic, the Thumby Color is not an emulator. It doesn’t have enough processing power to play retro games nor enough storage to hold hundreds of ROM files.

It’s powered by a 150MHz Raspberry Pi RP2350 processor that’s paired with just 16MB of storage — yes, that’s megabytes, not gigabytes. Not being able to play actual GBA games on the Thumby Color is probably a good thing, because its 0.85-inch square LCD screen has a resolution of just 128 x 128 pixels, which is less than half the resolution of the GBA.

Games are instead programmed in MicroPython. The Thumby Color will come with six open source titles pre-installed, including a first-person shooter called Monstra and a dungeon crawler called Thumbgeon 2. It will also come with all the monochromatic games that shipped with the original Thumby back in 2021, which was the company’s ultra-tiny clone of the original Game Boy.

The Thumby Color isn’t a handheld you’re going to want to play for hours on end, but it is one you can potentially sneak into any setting where you want a distraction — without trying to hide more conspicuous handhelds like the Steam Deck. It’s also a good excuse to learn a programming language as TinyCircuits has created tutorials and guides available for free through its website.

Image: TinyCircuits
The Thumby Color is available in several different colorways including transparent case options.

TinyCircuits has opted for a Kickstarter crowdfunding campaign to bring the Thumby Color to consumers. The original Thumby still sells for $42, but this new color screen version is available with a pledge of $49 for a purple version, with shipping expected as early as November. If you want a Thumby Color with a clear or transparent purple housing, the price jumps to $55, and if you want a Kickstarter exclusive Black version, it’s $75.

Keep in mind that there’s always the chance a crowdfunded product will not deliver all the features promised at the time of the campaign. It could also be significantly delayed, or never delivered at all, without refunds. TinyCircuits has successfully shipped several Kickstarted products before, so there’s slightly less risk here, but there’s no guarantee the product will materialize even with its funding goal already being met.

Read More 

Tesla owners won’t need to use their own adapter at Omni Port-capable stations. | Image: ChargePoint

ChargePoint has a new plug system for its charging stations that will make it easier for EV drivers to select the correct connecter for their electric vehicle. The new plug, called Omni Port, works by automatically installing an adapter (or excluding one) to the tip of the connector when you’re ready to charge, which means Tesla owners won’t need to bring out their adapter.
In a press release, ChargePoint CEO Rick Wilmer says Omni Port solves “the challenges associated with a multiple connector environment” and gives Tesla and non-Tesla drivers a “world-class” experience. As North American automakers transition to use the new NACS standard that’s based on Tesla’s connector, current drivers will continue needing access to J1772 Level 2 ports as well as CCS1 ports for DC fast charging. Yeah, charging is kind of a mess.

Image: ChargePoint
Now that’s a beefy CCS-to-NACS combo Omni Port for fast-charging a Tesla.

Omni Ports are shipping by the end of the year and will be included on new station hardware, including ones with support for 800-volt charging. Meanwhile, owners of ChargePoint’s commercial CP6000 and Express Plus Power Link 2000 stations can get an Omni Port retrofit, thanks to a modular design. However, don’t expect them to include the nearly extinct CHAdeMO port (although there might still be hope for Leaf owners).
ChargePoint’s Omni Port is like the reverse of Tesla’s limited “Magic Dock” Supercharger stations, where non-Teslas can unlock the charger and automatically have a CCS1 adapter tip installed. One advantage ChargePoint has, though, is the cable will reach your port no matter how you’re parked.

Tesla owners won’t need to use their own adapter at Omni Port-capable stations. | Image: ChargePoint

ChargePoint has a new plug system for its charging stations that will make it easier for EV drivers to select the correct connecter for their electric vehicle. The new plug, called Omni Port, works by automatically installing an adapter (or excluding one) to the tip of the connector when you’re ready to charge, which means Tesla owners won’t need to bring out their adapter.

In a press release, ChargePoint CEO Rick Wilmer says Omni Port solves “the challenges associated with a multiple connector environment” and gives Tesla and non-Tesla drivers a “world-class” experience. As North American automakers transition to use the new NACS standard that’s based on Tesla’s connector, current drivers will continue needing access to J1772 Level 2 ports as well as CCS1 ports for DC fast charging. Yeah, charging is kind of a mess.

Image: ChargePoint
Now that’s a beefy CCS-to-NACS combo Omni Port for fast-charging a Tesla.

Omni Ports are shipping by the end of the year and will be included on new station hardware, including ones with support for 800-volt charging. Meanwhile, owners of ChargePoint’s commercial CP6000 and Express Plus Power Link 2000 stations can get an Omni Port retrofit, thanks to a modular design. However, don’t expect them to include the nearly extinct CHAdeMO port (although there might still be hope for Leaf owners).

ChargePoint’s Omni Port is like the reverse of Tesla’s limited “Magic Dock” Supercharger stations, where non-Teslas can unlock the charger and automatically have a CCS1 adapter tip installed. One advantage ChargePoint has, though, is the cable will reach your port no matter how you’re parked.

Read More 

Image: OpenAI

OpenAI has released its GPT-4o System Card, a research document that outlines the safety measures and risk evaluations the startup conducted before releasing its latest model.
GPT-4o was launched publicly in May of this year. Before its debut, OpenAI used an external group of red teamers, or security experts trying to find weaknesses in a system, to find key risks in the model (which is a fairly standard practice). They examined risks like the possibility that GPT-4o would create unauthorized clones of someone’s voice, erotic and violent content, or chunks of reproduced copyrighted audio. Now, the results are being released.
According to OpenAI’s own framework, the researchers found GPT-4o to be of “medium” risk. The overall risk level was taken from the highest risk rating of four overall categories: cybersecurity, biological threats, persuasion, and model autonomy. All of these were deemed low risk except persuasion, where the researchers found some writing samples from GPT-4o could be better at swaying readers’ opinions than human-written text — although the model’s samples weren’t more persuasive overall.
An OpenAI spokesperson, Lindsay McCallum Rémy, told The Verge that the system card includes preparedness evaluations created by an internal team, alongside external testers listed on OpenAI’s website as Model Evaluation and Threat Research (METR) and Apollo Research, both of which build evaluations for AI systems.
This isn’t the first system card OpenAI has released; GPT-4, GPT-4 with vision, and DALL-E 3 were also similarly tested and the research was released. But OpenAI is releasing this system card at a pivotal time. The company has been fielding nonstop criticism of its safety standards, from its own employees to state senators. Only minutes before the release of GPT-4o’s system card, The Verge exclusively reported on an open letter from Sen. Elizabeth Warren (D-MA) and Rep. Lori Trahan (D-MA) that called for answers about how OpenAI handles whistleblowers and safety reviews. That letter outlines the many safety issues that have been called out publicly, including CEO Sam Altman’s brief ousting from the company in 2023 as a result of the board’s concerns and the departure of a safety executive, who claimed that “safety culture and processes have taken a backseat to shiny products.”
Moreover, the company is releasing a highly capable multimodal model just ahead of a US presidential election. There’s a clear potential risk of the model accidentally spreading misinformation or getting hijacked by malicious actors — even if OpenAI is hoping to highlight that the company is testing real-world scenarios to prevent misuse.
There have been plenty of calls for OpenAI to be more transparent, not just with the model’s training data (is it trained on YouTube?), but with its safety testing. In California, where OpenAI and many other leading AI labs are based, state Sen. Scott Wiener is working to pass a bill to regulate large language models, including restrictions that would hold companies legally accountable if their AI is used in harmful ways. If that bill is passed, OpenAI’s frontier models would have to comply with state-mandated risk assessments before making models available for public use. But the biggest takeaway from the GPT-4o System Card is that, despite the group of external red teamers and testers, a lot of this relies on OpenAI to evaluate itself.

Image: OpenAI

OpenAI has released its GPT-4o System Card, a research document that outlines the safety measures and risk evaluations the startup conducted before releasing its latest model.

GPT-4o was launched publicly in May of this year. Before its debut, OpenAI used an external group of red teamers, or security experts trying to find weaknesses in a system, to find key risks in the model (which is a fairly standard practice). They examined risks like the possibility that GPT-4o would create unauthorized clones of someone’s voice, erotic and violent content, or chunks of reproduced copyrighted audio. Now, the results are being released.

According to OpenAI’s own framework, the researchers found GPT-4o to be of “medium” risk. The overall risk level was taken from the highest risk rating of four overall categories: cybersecurity, biological threats, persuasion, and model autonomy. All of these were deemed low risk except persuasion, where the researchers found some writing samples from GPT-4o could be better at swaying readers’ opinions than human-written text — although the model’s samples weren’t more persuasive overall.

An OpenAI spokesperson, Lindsay McCallum Rémy, told The Verge that the system card includes preparedness evaluations created by an internal team, alongside external testers listed on OpenAI’s website as Model Evaluation and Threat Research (METR) and Apollo Research, both of which build evaluations for AI systems.

This isn’t the first system card OpenAI has released; GPT-4, GPT-4 with vision, and DALL-E 3 were also similarly tested and the research was released. But OpenAI is releasing this system card at a pivotal time. The company has been fielding nonstop criticism of its safety standards, from its own employees to state senators. Only minutes before the release of GPT-4o’s system card, The Verge exclusively reported on an open letter from Sen. Elizabeth Warren (D-MA) and Rep. Lori Trahan (D-MA) that called for answers about how OpenAI handles whistleblowers and safety reviews. That letter outlines the many safety issues that have been called out publicly, including CEO Sam Altman’s brief ousting from the company in 2023 as a result of the board’s concerns and the departure of a safety executive, who claimed that “safety culture and processes have taken a backseat to shiny products.”

Moreover, the company is releasing a highly capable multimodal model just ahead of a US presidential election. There’s a clear potential risk of the model accidentally spreading misinformation or getting hijacked by malicious actors — even if OpenAI is hoping to highlight that the company is testing real-world scenarios to prevent misuse.

There have been plenty of calls for OpenAI to be more transparent, not just with the model’s training data (is it trained on YouTube?), but with its safety testing. In California, where OpenAI and many other leading AI labs are based, state Sen. Scott Wiener is working to pass a bill to regulate large language models, including restrictions that would hold companies legally accountable if their AI is used in harmful ways. If that bill is passed, OpenAI’s frontier models would have to comply with state-mandated risk assessments before making models available for public use. But the biggest takeaway from the GPT-4o System Card is that, despite the group of external red teamers and testers, a lot of this relies on OpenAI to evaluate itself.

Read More 

Musk has chosen lawfare over free speech. | Image: Cath Virginia / The Verge, Getty Images

A group of advertisers, called the Global Alliance for Responsible Media (GARM), have “discontinued” activities, Business Insider reports, citing an email sent to members. The group said it had limited money, and was focused on fighting the antitrust lawsuit filed against it by X.
GARM was an initiative formed in 2019 by the World Federation of Advertisers (WFA); the idea was to stop advertising on social media platforms that don’t meet certain safety standards. In its lawsuit, X said that GARM “organized an advertiser boycott of Twitter.”
The lawsuit from the platform formerly known as Twitter was one of several odd legal actions coming from Elon Musk — ones that seem more focused on punishing anyone who doesn’t agree with him than based on any coherent legal theory. Musk has also sued OpenAI, a competitor to his X.AI, and Media Matters, to try to silence the non-profit.
“Go fuck yourself.”
Freedom of association is generally recognized as guaranteed by the First Amendment, and there’s precedent showing that boycotts are protected speech. But that’s not all that makes the GARM case odd. Musk himself has made comments that are hostile to advertisers. Last November, in an interview with Andrew Ross Sorkin, Musk said, “I hope they stop. Don’t advertise.”
He added, “If somebody is going to try to blackmail me with advertising, blackmail me with money, go fuck yourself. Go fuck yourself. Is that clear? I hope it is.”
When Musk bought Twitter, most of its revenue came from advertising — and that money dried up with his takeover. Perhaps the advertisers felt skittish about Musk’s “anything goes” approach to moderation — or perhaps it was Musk’s own posts, including one about “white pride.”
Though X’s case is hilariously thin, it’s accomplished its goal: defanging GARM. Taken together with some of Musk’s other suits, there’s a clear theme. if you oppose Musk, be ready to pay for expensive litigation.
X announced last month it was “excited” to rejoin GARM.

Musk has chosen lawfare over free speech. | Image: Cath Virginia / The Verge, Getty Images

A group of advertisers, called the Global Alliance for Responsible Media (GARM), have “discontinued” activities, Business Insider reports, citing an email sent to members. The group said it had limited money, and was focused on fighting the antitrust lawsuit filed against it by X.

GARM was an initiative formed in 2019 by the World Federation of Advertisers (WFA); the idea was to stop advertising on social media platforms that don’t meet certain safety standards. In its lawsuit, X said that GARM “organized an advertiser boycott of Twitter.”

The lawsuit from the platform formerly known as Twitter was one of several odd legal actions coming from Elon Musk — ones that seem more focused on punishing anyone who doesn’t agree with him than based on any coherent legal theory. Musk has also sued OpenAI, a competitor to his X.AI, and Media Matters, to try to silence the non-profit.

Freedom of association is generally recognized as guaranteed by the First Amendment, and there’s precedent showing that boycotts are protected speech. But that’s not all that makes the GARM case odd. Musk himself has made comments that are hostile to advertisers. Last November, in an interview with Andrew Ross Sorkin, Musk said, “I hope they stop. Don’t advertise.”

He added, “If somebody is going to try to blackmail me with advertising, blackmail me with money, go fuck yourself. Go fuck yourself. Is that clear? I hope it is.”

When Musk bought Twitter, most of its revenue came from advertising — and that money dried up with his takeover. Perhaps the advertisers felt skittish about Musk’s “anything goes” approach to moderation — or perhaps it was Musk’s own posts, including one about “white pride.”

Though X’s case is hilariously thin, it’s accomplished its goal: defanging GARM. Taken together with some of Musk’s other suits, there’s a clear theme. if you oppose Musk, be ready to pay for expensive litigation.

X announced last month it was “excited” to rejoin GARM.

Read More 

Nanoleaf’s Essentials Matter line, which includes an A19 bulb and light strip, will be one of the first Matter products to work with Apple Home’s Adaptive Lighting feature. | Photo by Jennifer Pattison Tuohy / The Verge

A major feature of Apple Home will soon work with Matter devices. With iOS 18, Apple is enabling Adaptive Lighting for smart lights connected to its smart home platform through Matter, the interoperability standard developed by Apple, Amazon, and Google. The feature, which automatically adjusts the color temperature of smart lights throughout the day, currently only works with a few brands. Adding support for Matter bulbs means it should work with many more.

Gimmy Chu, CEO of Nanoleaf, confirmed to The Verge that his company is currently testing the feature on its line of Essentials Matter lights. “We are one of the partners that are part of the launch of Adaptive Lighting on Matter products,” he said in an interview. “For the past several months we’ve been doing extensive testing to ensure that the customers will have a great experience.”
This testing has led to Adaptive Lighting for Nanoleaf’s Matter products appearing in the iOS 18 betas, which is how The Verge first saw it. Users on Reddit in the betas reported seeing the toggle to enable Adaptive Lighting on their Nanoleaf smart bulbs.

Screenshot by Wes Davis / The Verge
Adaptive Lighting is a feature you can enable in the Apple Home app with compatible bulbs, of which there should soon be more.

Adaptive Lighting is popular among HomeKit users. Its ability to automatically adjust your lighting color temperature throughout the day, turning lights from cool to warm to make your home feel bright and refreshing or welcoming and cozy, is an easy quality-of-life upgrade for your lights. It’s been part of Apple Home since iOS 14, but uptake among manufacturers has been slow — Nanoleaf and Philips Hue were among the first, and Eve and Aqara also have compatible products. But that’s about it.
However, when Hue and Nanoleaf released Matter versions of their products (Hue by an upgrade to its bridge), Adaptive Lighting wasn’t available. Understandably, people were frustrated that the Matter products had less functionality than the existing products.
Chu explained that as much as Nanoleaf wanted to have Adaptive Lighting on its new Matter products, it wasn’t possible because Apple hadn’t enabled support for the feature in smart lights connected to its platform through Matter rather than HomeKit. (Hue users could keep Adaptive Lighting by not upgrading their bridge to Matter.)
But now, when iOS 18 arrives this fall, all of Nanoleaf’s Matter lights will work with Adaptive Lighting in Apple Home. The Verge reached out to Philips Hue to ask if it’s planning to enable the feature on its Matter bridge; as of publication, we have not heard back.

While this is good news for Apple Home users with Nanoleaf products (and likely Hue ones, too), it’s also exciting for those with other brands of smart lights. “I believe that the vision for Adaptive Lighting is to be open to all partners that work with the Apple Home ecosystem via Matter,” said Chu.
That means Matter-compatible lighting products, including those from companies like Govee, Lifx, Wiz, and Linkind, could also work with the feature through Apple Home. The more compatible products there are, the easier (and cheaper) it will be to use Adaptive Lighting throughout your home.
Adaptive Lighting is an Apple Home exclusive feature and is not part of the Matter spec. This means it won’t work in other Matter-compatible ecosystems, like Amazon Alexa or Google Home. However, the Connectivity Standards Alliance, the organization that runs Matter, has previously said that more advanced lighting features similar to Adaptive Lighting and the dynamic lighting effects companies like Nanoleaf and Govee use are on its roadmap. If that happens, then those would work across all compatible platforms.

Nanoleaf’s Essentials Matter line, which includes an A19 bulb and light strip, will be one of the first Matter products to work with Apple Home’s Adaptive Lighting feature. | Photo by Jennifer Pattison Tuohy / The Verge

A major feature of Apple Home will soon work with Matter devices. With iOS 18, Apple is enabling Adaptive Lighting for smart lights connected to its smart home platform through Matter, the interoperability standard developed by Apple, Amazon, and Google. The feature, which automatically adjusts the color temperature of smart lights throughout the day, currently only works with a few brands. Adding support for Matter bulbs means it should work with many more.

Gimmy Chu, CEO of Nanoleaf, confirmed to The Verge that his company is currently testing the feature on its line of Essentials Matter lights. “We are one of the partners that are part of the launch of Adaptive Lighting on Matter products,” he said in an interview. “For the past several months we’ve been doing extensive testing to ensure that the customers will have a great experience.”

This testing has led to Adaptive Lighting for Nanoleaf’s Matter products appearing in the iOS 18 betas, which is how The Verge first saw it. Users on Reddit in the betas reported seeing the toggle to enable Adaptive Lighting on their Nanoleaf smart bulbs.

Adaptive Lighting is popular among HomeKit users. Its ability to automatically adjust your lighting color temperature throughout the day, turning lights from cool to warm to make your home feel bright and refreshing or welcoming and cozy, is an easy quality-of-life upgrade for your lights. It’s been part of Apple Home since iOS 14, but uptake among manufacturers has been slow — Nanoleaf and Philips Hue were among the first, and Eve and Aqara also have compatible products. But that’s about it.

However, when Hue and Nanoleaf released Matter versions of their products (Hue by an upgrade to its bridge), Adaptive Lighting wasn’t available. Understandably, people were frustrated that the Matter products had less functionality than the existing products.

Chu explained that as much as Nanoleaf wanted to have Adaptive Lighting on its new Matter products, it wasn’t possible because Apple hadn’t enabled support for the feature in smart lights connected to its platform through Matter rather than HomeKit. (Hue users could keep Adaptive Lighting by not upgrading their bridge to Matter.)

But now, when iOS 18 arrives this fall, all of Nanoleaf’s Matter lights will work with Adaptive Lighting in Apple Home. The Verge reached out to Philips Hue to ask if it’s planning to enable the feature on its Matter bridge; as of publication, we have not heard back.

While this is good news for Apple Home users with Nanoleaf products (and likely Hue ones, too), it’s also exciting for those with other brands of smart lights. “I believe that the vision for Adaptive Lighting is to be open to all partners that work with the Apple Home ecosystem via Matter,” said Chu.

That means Matter-compatible lighting products, including those from companies like Govee, Lifx, Wiz, and Linkind, could also work with the feature through Apple Home. The more compatible products there are, the easier (and cheaper) it will be to use Adaptive Lighting throughout your home.

Adaptive Lighting is an Apple Home exclusive feature and is not part of the Matter spec. This means it won’t work in other Matter-compatible ecosystems, like Amazon Alexa or Google Home. However, the Connectivity Standards Alliance, the organization that runs Matter, has previously said that more advanced lighting features similar to Adaptive Lighting and the dynamic lighting effects companies like Nanoleaf and Govee use are on its roadmap. If that happens, then those would work across all compatible platforms.

Read More