Grammarly announced today that it’s (unsurprisingly) diving into the generative AI fray. GrammarlyGo is an upcoming set of auto-composition features to help the AI proofreading software keep up with the many companies adding the ChatGPT API (or different generative AI backends) to their products.

GrammarlyGo can use context like voice, style, purpose and where you’re writing to determine its approach. So, for example, it can spit out email replies, shorten passages, rewrite them for tone and clarity, brainstorm or choose from one-click prompts — all while adhering to your company’s voice or other provided context. In addition, since Grammarly’s desktop service can pop up in any text field on your computer, its generative writing could be slightly more convenient than competitors (like Notion or Gmail’s Smart Compose) that require you to visit an app or website. The company says GrammarlyGo will be enabled by default for individuals, and you can toggle it in settings.

Grammarly justifies the feature’s existence by saying most people’s writing can be better and faster. “Individuals today spend too much time trying to communicate in the right way, while poor communication is draining business productivity and performance,” the company wrote in an announcement post. “GrammarlyGO will address this problem by quickly generating highly relevant text with an understanding of personal voice and brand style, context, and intent — saving people and businesses time while accounting for their unique needs.” I’m unsure what to make of a world where nobody writes anything but prompts for machines, but that increasingly appears to be where we’re headed.

GrammarlyGo will be included with the Grammarly service at no extra cost. The company says it will soon add the AI writing feature to its Premium, Business, Education and Developer plans — and free plans “in select markets.” The GrammarlyGo beta will begin rolling out in April.

As a white man in America with no discernible regional accent, I can simply assume that modern consumer technologies — virtual assistants like Siri, Alexa or Assistant, and my phones’ camera — will work seamlessly out of the box. I assume this because, well, they do. That’s namely because the nerds who design and program these devices overwhelmingly both look and sound just like me — if even a little whiter. Folks with more melanin in their skin and extra twang on their tongue don’t enjoy that same privilege.

Tomorrow’s chatbots and visual AIs will only serve to exacerbate this bias unless steps are taken today to ensure a benchmark standard of fairness and equitable behavior from these systems. To address that issue, Meta AI researchers developed and released the Casual Conversations dataset in 2021, designed to “help researchers evaluate their computer vision and audio models for accuracy across a diverse set of age, genders, apparent skin tones and ambient lighting conditions.” On Thursday, the company unveiled Casual Conversations v2, which promises even more granular classification categories than its predecessor.

The original CC dataset included 45,000 videos from more than 3,000 paid subjects across age, gender, apparent skin tone and lighting conditions. These videos are designed to be used by other AI researchers, specifically those working with generative AIs like ChatGPT or visual AIs like those used in social media filters and facial recognition features, to help them ensure that their creations behave the same whether the user looks like Anya Taylor-Joy or Lupita Nyong’o, whether they sound like Colin Firth or Colin Quinn.

Since Casual Conversations first debuted two years ago, Meta has worked “in consultation with internal experts in fields such as civil rights,” according to Tuesday’s announcement, to expand and improve upon the dataset. Professor Pascale Fung, director of the Centre for AI Research, as well as other researchers from Hong Kong University of Science and Technology, participated in the literature review of government and industry data to establish the new annotation categories.

Version 2 now includes 11 categories (seven self-reported and four researcher-annotated) and 26,467 video monologues recorded by nearly 5,600 subjects in seven countries — Brazil, India, Indonesia, Mexico, Vietnam, Philippines and the US. While there aren’t as many individual videos in the new dataset, they are far more heavily annotated. As Meta points out, the first iteration only had a handful of categories: “age, three subcategories of gender (female, male, and other), apparent skin tone and ambient lighting,” according to the Thursday blog post.

“To increase nondiscrimination, fairness, and safety in AI, it’s important to have inclusive data and diversity within the data categories so researchers can better assess how well a specific model or AI-powered product is working for different demographic groups,” Roy Austin, Vice President and Deputy General Counsel for Civil Rights at Meta, said in the release. “This dataset has an important role in ensuring the technology we build has equity in mind for all from the outset."

As with most all of its public AI research to date, Meta is releasing Casual Conversations v2 as an open source dataset for anyone to use and expand upon — perhaps to include markers such as “disability, accent, dialect, location, and recording setup,” as the company hinted at on Thursday.

Spoiler warning: The following article discusses the Star Trek: Picard episodes “Seventeen Seconds” and “No Win Scenario.”

I was away last week and so didn’t write up last week’s utterly inessential episode of Picard, although if you’re curious, Darren Mooney at The Escapist’s feelings mirror my own. The whole affair only makes sense if you assume the whole crew had taken a big dose of idiot pills an hour before the episode began, and wish you’d had some as well. This week, there’s a lot more to say, especially since it’s the first episode this season that feels even remotely fun. And while the setup for all of these conflicts was pretty flimsy, their resolutions are all very enjoyable to watch.

After the Shrike uses its Portal gun to encourage the Titan to shoot itself in the back, the ship drifts into a gravity well and certain doom. It’s here we get one of the worst tell-not-show moments in the series so far as the bridge crew state, and then restate (and then re-restate) the situation the ship finds itself in. Loss of power, is it? Sinking into a gravity well, is it? We don’t have enough power to get out of the gravity well, is it? Is that because we’ve not got much power, is it? And because of the gravity well? Sorry, not following, can you explain it to me like I’m five, please? It doesn’t help that while the decision to take the shields offline is set up as some big, dangerous decision, it’s never mentioned again.

It’s here that Picard, opting to get his affairs in order, chooses to sit down for a touching moment with Jack. And they choose to do so on the holodeck, in the Ten Forward bar that’s been turning up time and again this run. Picard says that holodecks are hooked up to their own power supply because it’s better to die in comfort than use that power to survive. I think we can all tell that it’s a cover for either a production or a budget issue that meant they had to re-use the set. (Picard’s first season did the same, endlessly going back to the chateau office on La Sirena’s holodeck.) And, again, the two-hander between Patrick Stewart and Ed Speelers is great.

Also, remember when a broken leg wouldn’t require much more than a quick waft of a med bay doodad and you’d be good as new? Not in Picard, where Shaw storms into Picard’s heart to heart, apparently full of painkillers, and reveals why he’s so angry at both Seven and Picard. It turns out that, like a big chunk of Starfleet, he’s a survivor from Wolf 359! And if this wasn’t another thing that was explored pretty well by Next Generation and Deep Space Nine, it might carry a bit more dramatic weight. It also feels like the start of a pre-ordained redemption arc for Todd Stashwick’s Captain Shaw as the powers that be use Picard as a way to backdoor pilot a Titan-A spin-off series. (You can imagine the pitch: We’ve got the cast, and the sets are already built, it’s practically free content if you greenlight another ten!) And, to be fair, Todd Stashwick’s such a charismatic actor that you could feel him straining to play someone more unlikeable at the start of the series.

From here until the conclusion, however, the tone starts to lighten, and Beverley’s realization that the nebula is a space lifeform gives everyone a sense of purpose. You see, the nebula is undergoing the exact same contractions that a pregnant person would when they’re in labor. The moment that Riker stops pointlessly objecting to the plan of riding the contractions out of the gravity well, it suddenly feels like we’re watching Star Trek again. The gang works together, Seven and Shaw successfully lure out the changeling infiltrator, and they even have a nice spot of payback for the Shrike as Riker orders the Titan to lock onto a massive asteroid, dragging it behind long enough for it to smash the enemy vessel out of contention long enough for them to escape.

And that’s not even the best bit, because there’s also the wonderful B-Story of Picard dealing with his adoring fans while in (the real) Ten Forward five years previously. Patrick Stewart Picard is accosted by a bunch of fans cadets who ask him to regale them with stories of his time on Star Trek the Enterprise. All the while, unseen, Jack lingers in the background, listening to Picard as he builds out his myth and his legacy while minimizing any reference to his own family. When Picard closes the gathering by saying that Starfleet is his family, it’s both an unwittingly hollow indictment of Picard’s own life (his co-workers are his only friends, oof) and an unwitting rejection of the son he could have grown to know years previously. This, my friends, is a great moment, full of depth and pathos and I just wish that it hadn’t taken this long to get here.

Speaking of which, Paramount recently punted Star Trek: Discovery’s fifth season to 2024, adding it would now be that show’s final run. With news that the studio is looking to tighten its belt in order to milk some actual profit from its streaming service, fans are feeling antsy. After all, Trek shows aren’t cheap to make, and it’s not clear how much crossover appeal these shows have. Despite David Stapf’s 2018 promise that we’d have “a Star Trek something on all the time,” there’s a worry that it’s closing time at the all-you-can-eat buffet. With Discovery and Picard on the outs, and no sign (yet) that Lower Decks and Prodigy will get renewals, we could go from five shows to three, or one, in no time at all. But, based on the merits of some of what’s been released under the Trek brand of late, would that really be such a bad thing?

After all, these four episodes of Picard form little more than an extended prologue for the rest of the run. It’s taken the better part of four hours to establish the broad outline of the plot as well as the main antagonists, and the stakes at hand. Even then, we’ve still not encountered more than half of the Next Generation returnees who formed such a key part of the marketing. A prologue that I would have enjoyed a hell of a lot more if it had been compressed down to closer to ninety minutes. Imagine if, rather than filling out a corporate-mandated ten-week block each year, the format was designed to suit the story being told at each time. On the merits of the last four weeks alone, fewer episodes of higher quality Trek would be infinitely preferable.

At CES, Roku announced that it'll finally be building its own Roku TVs, in addition to working with partners like TCL and Hisense. Starting today, you can pick up one of those sets at Best Buy's stores and website. Roku is offering two series of TVs, both of which are powered by its streaming platform: Roku Select sets ranging from 24 to 75-inches, and Roku Plus televisions coming in at 55-, 65-, and 75-inches.

We're still waiting for a full pricing breakdown, but Roku says Select TVs will start at $149 (with a potential drop to $120 this summer), and the Plus models will start below $500. That's in-line with similar offerings from TCL and other partners. At CES, Roku's VP of retail strategy, Chris Larson, said the company is aiming to deliver features partners don't always include, like voice-enabled remotes, as well as automatic brightness adjustment (on the Premiere sets).

Roku isn't saying much about its Select TVs, perhaps because there's not much to say. They'll include HD and 4K models (there's no need to go 4K on very small TVs, after all), as well as its standard voice remote. Like most bargain TVs, they'll also rely on edge dimming backlights — so expect to see gray-like blacks and spotty lighting. The Roku Plus sets are far more advanced: They'll include QLED panels, local backlight dimming for better contrast, 4K Dolby Vision and Dolby Atmos-approved speakers. They'll also get Roku's rechargeable voice remote.

While Roku told us that it's not trying to compete with its longtime partners, it'll be interesting to see how those companies respond. By making voice remotes standard, perhaps TCL and others will follow suit. The company also unveiled a Roku OS update today, which will include local news and support for premium subscriptions on Roku's live channel guide.

If you want to change the appearance of subtitles and closed captions on Netflix — perhaps to switch font sizes or to add a shadow that would make them more readable — you'd have to access your account on a browser. The process is a lot more complicated than it needs to be if you primarily stream on TV. In fact, I've never changed what mine looks like all these years, even if I had previously wanted to. Now, according to TechCrunch, the streaming giant has finally started rolling out the capability to customize subtitle and closed caption text within the TV app itself.

Based on the publication's report, the feature looks and works similar to Amazon Prime Video TV app's subtitle adjuster. It allows you to choose between small, medium and large text, as well as between four styles with and without drop shadows. One option is yellow text on a black background for greater contrast if you need it. On the service's updated help page, it said you can access those controls by going to Settings at the bottom of the screen while a show is playing and then choosing subtitle Size and Style. Take note that it's not possible to change text appearance for Arabic, Hebrew, Japanese, Korean, Romanian, Thai and Traditional Chinese at the moment. 

More and more people are using subtitles in recent years to hear dialogue when sounds are unbalanced and voices seem to be lower in volume than the background, or to watch shows in a foreign language. It's not just viewers with hard of hearing that use them anymore. And while the controls for the TV app offer fewer customization options than Netflix's website, they make changing up subtitles' appearance a lot more accessible. 

It’s easy to be cynical about a lot of things, but then you see a picture of clouds taken from the soil of another planet, and you remember what a great gift human ingenuity can be. For the first time, a Mars probe has taken a clear image of sunbeams seen from the Martian surface. Curiosity rover, during a survey of twilight cloud formations, took a shot clearly showing crepuscular rays emerging from the heavens.

Researchers say this cloud formation is higher in the atmosphere than normal, and is made of dry ice rather than water ice. Other treats from Curiosity’s current survey include sight of an iridescent cloud shaped like a feather. This, too, is helping teams here on Earth get a better sense of the Martian atmosphere, not to mention giving us a lovely stream of new desktop backgrounds.

– Dan Cooper

The Morning After isn’t just a newsletter – it’s also a daily podcast. Get our daily audio briefings, Monday through Friday, by subscribing right here.

The biggest stories you might have missed

• Sony's PS5 update with Discord voice chat begins rolling out to all

• The best wireless chargers in 2023

• Watch Relativity Space try to launch a 3D-printed rocket into orbit at 1 PM ET (update: scrubbed)

• Microsoft debuts a new 'velocity green' Xbox controller

• The most interesting foldables, rollables and demos at MWC 2023

• Microsoft Edge can now use AMD and NVIDIA GPUs to upscale web videos

'Starfield' has been delayed to September 6th

We’ll learn more about the ambitious RPG in June.

Bethesda

Starfield, Bethesda Studios’ ambitious sci-fi RPG, has once again been delayed, this time to September 6th. That may be a pain to fans who’ve been waiting since the last pledged release date of November 11th, 2022, to get their hands on the title. In compensation, the studio will release a lengthy deep-dive on June 11th, immediately after Microsoft and Bethesda’s own Xbox summer showcase ahead of E3.

Continue Reading.

VW's connected emergency service is free for 5 years after botched carjacking response

It refused police requests to activate dormant tracking features to help save a child.

Volkswagen is offering its Car-Net connected emergency service free to a number of 2020 to 2023 vehicles after a high-profile bungle. An Atlas SUV was carjacked, with the thief not realizing a child was still inside, but VW staff refused requests made by police officers to activate remote tracking. Mercifully, the child was found safe and well, but as part of a high-profile mea culpa, VW will offer five years’ service free to all qualifying models.

Continue Reading.

Ring’s latest doorbell offers head-to-toe views of your visitor

The Battery Doorbell Plus is the first battery-powered machine to get the feature.

Ring

Ring’s brought one of the best features from its Pro 2 video doorbell to its range of battery powered units. The new Battery Doorbell Plus gets the same 150 x 150-degree field of vision and 1536p video as the flagship. That gives you a proper view of your doorstep, letting you keep an eye on packages dumped there when you’re out.

Continue Reading.

Google One's VPN will soon be available to all subscribers

That includes base-tier subscribers paying as little as $2 per month.

Google’s One subscription plan offers cloud storage and backup, extra support and better offers on Google products. It also has a VPN, but that’s only on offer to people paying for the Premium tier with 2TB of storage. Now, however, the search giant is opening up its VPN for all users and will even bring the feature to base-tier users paying as little as $2 a month.

Continue Reading.

Bowers & Wilkins Pi7 S2 review: Excellent sound only goes so far

The whole package lacks polish, especially at this sort of price.

Billy Steele

Our audio expert Billy Steele has reviewed Bowers & Wilkins’ new Pi7 S2 true wireless earbuds, and it makes for great reading. Unless you’re Bowers or Wilkins. He takes the company to task for not pairing the admittedly great sound with a package to match it. Lackluster noise cancellation, poor battery life and limited customization paired with a price far in excess of its better, cheaper competitors means you might be better off looking elsewhere for your next pair of earbuds.

Continue Reading.

Global shutter sensors with no skew or distortion have been promised as the future of cameras for years now, but so far only a handful of products with that tech have made it to market. Now, Raspberry Pi is offering a 1.6-megapixel global shutter camera module to hobbyists for $50, providing a platform for machine vision, hobbyist shooting and more. 

The Raspberry Pi Global Shutter Camera uses a 6.3mm Sony IMX296 sensor, and requires a Raspberry Pi board with a CSI camera connector. Like other global shutter sensors, it works by pairing each pixel with an analog storage element, so that light signals can be captured and stored by all pixels simultaneously. 

By comparison, regular CMOS sensors read and store the light captured by pixels from top to bottom and left to right. That can cause diagonal skew on fast moving subjects, or very weird distortion on rotating objects like propellers. The video below shows the difference with plucked guitar strings. 

By eliminating those issues, the new camera allows for distortion free capture of things like sports or fast-moving industrial processes. The relatively low resolution isn't a problem, the company says, as video is usually downsampled before being fed into machine vision systems anyway. It uses the same C/CS lens mount as Raspberry Pi's 12-megapixel High Quality Camera, so you can attach 6mm CS‑mount and 16mm C-mount CGL lenses offered through the company's reseller partners. 

If you're interested but worried about delays, Raspberry Pi recently posted that it has been working on resolving supply chain issues. "We expect supply to recover to pre-pandemic levels in the second quarter of 2023, and to be unlimited in the second half of the year," it said in a December blog post. The Global Shutter Camera is now available to purchase for $50. 

The Oversight Board has agreed to review a case related to Meta’s handling of election content in Brazil. In a statement, the board said they planned to scrutinize the social network’s policies surrounding election content in “high-risk” areas.

The case stems from a user who posted a video in early January calling for people to “besiege” Brazil’s congress following the election of President Lula da Silva. The video also featured clips of a speech from a Brazilian general, who called for people to go into the streets and government buildings. The video was reported seven times by four different users, according to the board, but remained on Facebook even after it was reviewed by five separate moderators. Meta later opted to remove the post and issue a “strike” to the person who had originally posted it, following the Oversight Board’s decision to review the case.

Though the case is related to Brazil’s most recent presidential election, the board’s recommendations could have a more-far reaching impact. “The Board selected this case to examine how Meta moderates election-related content, and how it is applying its Crisis Policy Protocol in a designated ‘temporary high-risk location,’” the group wrote in a statement.

As the board points out, Meta’s “Crisis Policy Protocol,” is a central aspect of the case. The protocol, which was created after the Oversight board weighed in on the suspension of Donald Trump, allows Meta to respond to situations when there is a risk of “imminent harm” either offline or online. So any recommendations that address that policy could end up affecting election-related content around the world, not just in Brazil.

However, that outcome is still months away. For now, the Oversight Board is asking for public feedback on various issues associated with the case before it makes recommendations to Meta. The company will then have 60 days to respond, though, as usual, Meta is not required to adopt policy changes suggested by the board.

Bing has crossed 100 million daily active users a month after the launch of its chatbot AI, according to Yusuf Mehdi, Microsoft's VP for Modern Life, Search and Devices. He said the company is fully aware that it's still just "a small, low, single digit share player," but hey, there was a time when Bing wasn't even a part of the conversation. Now, after the tech giant released its next-gen version, even those who haven't used it in the past are relying on it for their searches: Mehdi noted that one-third of Bing's daily active users are new to the search engine.

"We see this appeal of the new Bing as a validation of our view that search is due for a reinvention and of the unique value proposition of combining Search + Answers + Chat + Creation in one experience," the VP said. 

In addition to seeing a boost in numbers, Microsoft is also apparently enjoying a growth in engagement, with more people conducting more searches. The company credits two factors for that particular victory, the first being Edge's growth in usage, most likely aided by the addition of Bing's chat AI as a new feature. It also said that the introduction of its Prometheus AI model made Bing's search results more relevant, so people have been using — or at least trying out — the search engine more. 

Apparently, around one third of Bing's daily preview users have been using its chat AI for their queries every day. On average, Microsoft is seeing three chats per session, with over 45 million chats since it introduced the new Bing. Further, in 15 percent of all chat sessions, people have been using Bing to generate new content. The launch of Bing's AI chatbot on mobile has propelled the search engine to a new level of popularity, as well, and has led to a six-time increase in daily active users from before it became available. 

By integrating an AI chatbot into Bing earlier this year, Microsoft gave its search engine the weapon it needs to be able to compete against Google. That said, Google does have plans to launch a chatbot of its own and introduced a chat AI called Bard last month. Bard spouted misinformation during its unofficial debut, but Google is working with employees to improve the chatbot's responses before it becomes available.

Nothing, Forever, an AI-generated livestream inspired by Seinfeld is back on Twitch after being taken offline for an anti-LGBTQ outburst, Gizmodo has reported. After launching in December, the show (from Mismatch Media) became internet-famous for its Seinfeld-adjacent plots, janky '90s-era video game style animation and terrible (though coherent) dialogue. However, it was pulled over a month ago after violating Twitch's conduct policy banning hate speech around sexual and gender identity.

In the most notorious incident, the AI lead character ("Larry"), went on an offensive rant. "I’m thinking about doing a bit about how being transgender is actually a mental illness. Or how all liberals are secretly gay and want to impose their will on everyone. Or something about how transgender people are ruining the fabric of society. But no one is laughing, so I’m going to stop."

Shortly afterwards, Mismatch Media cofounder Skyler Hartle explained that the problem started when its OpenAI GPT-3 Davinci model stopped working correctly. The team switched to Davinci's predecessor Curie, believing that OpenAI's content moderation was still active — which was apparently not the case. The offensive outbursts started shortly afterwards.

"We mistakenly believed that we were leveraging OpenAI’s content moderation system for their text generation models. We are working now to implement OpenAI’s content moderation API (it’s a tool we can use to verify the safeness of the content) before we go live again, and investigating secondary content moderation systems as redundancies," the team said. 

Problems of moderation and offensive content have plagued AI chatbots since they arrived. Microsoft has experienced it twice, first with its Tay chatbot that turned racist, and later with the Bing Chat search assistant powered by OpenAI tech. The latter was pulled for a time after it started insulting users and outputting incorrect information, while insisting it was right. 

Nothing, Forever seems to be working as before, with the same laugh track, engaged viewers (around 3,500 currently) and complete disregard for collision detection — but no offensive outbursts so far. While nothing in the show makes any sense, the fact that AI can generate all the elements in real time is impressive.