verge-rss

Picsart’s AI image generator will be available some time later this year. | Image: Picsart

Image and video-editing platform Picsart has teamed up with Getty Images to launch a “responsible, commercially-safe” AI image generator that’s trained exclusively on the stock image platform’s licensed content.
Picsart says its AI lab is building a custom model from scratch that will power the tool, allowing the platform’s paid subscribers to generate images with full commercial rights — which aims to address long-standing concerns regarding AI-generated potentially content violating copyright laws. The Picsart / Getty Images generator tool is expected to launch “later this year,” and will be accessible through Picsart’s API services.
This collaborative project shares similarities with Adobe’s Firefly AI model, which initially launched as a prompt-based image generation tool in Photoshop last year before being integrated more widely across Creative Cloud apps. That “commercially safe” generator is also trained on stock images (specifically Adobe’s own Stock library), alongside content that’s openly licensed or out of copyright — but that training data isn’t squeaky clean, and its users don’t all trust how the company is training its model.

Image: Picsart
Picsart’s blog contained these examples supposedly generated using its upcoming AI model.

Getty has also worked on other commercially-focused AI products before, having announced similar partnerships with Bria AI and Runway, and collaborating with Nvidia to launch its own “Generative AI by Getty Images” tool trained on its vast library of licensed images. Given Adobe has rolled out its Firefly model into extremely popular apps like Photoshop, Illustrator, Lightroom, and Express, however, it may be difficult to convince creatives to jump ship to Picsart’s upcoming offering.

Picsart’s AI image generator will be available some time later this year. | Image: Picsart

Image and video-editing platform Picsart has teamed up with Getty Images to launch a “responsible, commercially-safe” AI image generator that’s trained exclusively on the stock image platform’s licensed content.

Picsart says its AI lab is building a custom model from scratch that will power the tool, allowing the platform’s paid subscribers to generate images with full commercial rights — which aims to address long-standing concerns regarding AI-generated potentially content violating copyright laws. The Picsart / Getty Images generator tool is expected to launch “later this year,” and will be accessible through Picsart’s API services.

This collaborative project shares similarities with Adobe’s Firefly AI model, which initially launched as a prompt-based image generation tool in Photoshop last year before being integrated more widely across Creative Cloud apps. That “commercially safe” generator is also trained on stock images (specifically Adobe’s own Stock library), alongside content that’s openly licensed or out of copyright — but that training data isn’t squeaky clean, and its users don’t all trust how the company is training its model.

Image: Picsart
Picsart’s blog contained these examples supposedly generated using its upcoming AI model.

Getty has also worked on other commercially-focused AI products before, having announced similar partnerships with Bria AI and Runway, and collaborating with Nvidia to launch its own “Generative AI by Getty Images” tool trained on its vast library of licensed images. Given Adobe has rolled out its Firefly model into extremely popular apps like Photoshop, Illustrator, Lightroom, and Express, however, it may be difficult to convince creatives to jump ship to Picsart’s upcoming offering.

Read More 

Why read the article when you can just AI generate takeaways? | Image: Yahoo

Artifact is dead, long live Yahoo’s version of Artifact. The architecture behind Artifact, the news aggregation app built by Instagram co-founders Kevin Systrom and Mike Krieger, will live on inside the body of a brand-new Yahoo News app.
Available to download today on iOS and Android, the new Yahoo News app brings an AI-powered personalized news feed for users based on their interests, while a feature called Key Takeaways can give a bullet summary of a news article when a reader is feeling TL;DR.
Other features of the Yahoo News app include Top Stories, which picks up on trending stories for users to read and will soon include key takeaway summaries. You can block stories with undesired keywords, as well as filter out certain publishers to your preference. And just like Artifact, Yahoo News also lets you flag content like clickbaity headlines, then lets AI rewrite them to something better.
Yahoo purchased Artifact in April for an undisclosed amount, saving it from total demise. Artifact had sort of lost itself in the final stretch of its life after it had tried making it into a Twitter-like social media platform — although its original goal was to be like a “TikTok for Text.” Yahoo hasn’t said it’s bringing all of the social features of Artifact back, but the new app allows you to share article excerpts with your friends. It also has a gamified “streak” feature and badges you can show to others for being a heavy reader.
Yahoo is also taking some of what it’s building in the News app to the Yahoo News online homepage. Starting today, the website has a new layout that highlights top news, gives personalized recommendations, and shows trending topics. The new homepage experience is opt-in.

Why read the article when you can just AI generate takeaways? | Image: Yahoo

Artifact is dead, long live Yahoo’s version of Artifact. The architecture behind Artifact, the news aggregation app built by Instagram co-founders Kevin Systrom and Mike Krieger, will live on inside the body of a brand-new Yahoo News app.

Available to download today on iOS and Android, the new Yahoo News app brings an AI-powered personalized news feed for users based on their interests, while a feature called Key Takeaways can give a bullet summary of a news article when a reader is feeling TL;DR.

Other features of the Yahoo News app include Top Stories, which picks up on trending stories for users to read and will soon include key takeaway summaries. You can block stories with undesired keywords, as well as filter out certain publishers to your preference. And just like Artifact, Yahoo News also lets you flag content like clickbaity headlines, then lets AI rewrite them to something better.

Yahoo purchased Artifact in April for an undisclosed amount, saving it from total demise. Artifact had sort of lost itself in the final stretch of its life after it had tried making it into a Twitter-like social media platform — although its original goal was to be like a “TikTok for Text.” Yahoo hasn’t said it’s bringing all of the social features of Artifact back, but the new app allows you to share article excerpts with your friends. It also has a gamified “streak” feature and badges you can show to others for being a heavy reader.

Yahoo is also taking some of what it’s building in the News app to the Yahoo News online homepage. Starting today, the website has a new layout that highlights top news, gives personalized recommendations, and shows trending topics. The new homepage experience is opt-in.

Read More 

Image: Microsoft

Microsoft is planning to launch its new Copilot Plus PCs next week without its controversial Recall feature that screenshots everything you do on these new laptops. The software maker is holding back Recall so it can test it with the Windows Insider program, after originally promising to ship Recall as an opt-in feature with additional security improvements.
“We are adjusting the release model for Recall to leverage the expertise of the Windows Insider community to ensure the experience meets our high standards for quality and security,” says Microsoft in an updated blog post. “When Recall (preview) becomes available in the Windows Insider Program, we will publish a blog post with details on how to get the preview.”
This means that Recall won’t even be available initially to Windows Insiders or anyone who buys a Copilot Plus PC. I wrote in Notepad earlier today that Windows engineers were scrambling to get the security improvements tested and implemented in time for the June 18th launch date of Copilot Plus PCs. Now, Microsoft is essentially admitting here that it needs more time to test Recall’s security improvements.

GIF: Microsoft
The Windows Recall feature.

Microsoft first unveiled the Recall feature as part of its upcoming Copilot Plus PCs last month, but since then, privacy advocates and security experts have been warning that Recall could be a “disaster” for cybersecurity without changes. Microsoft committed to three major updates to Recall last week, including making the AI-powered feature an opt-in experience instead of on by default, encrypting the database, and authenticating through Windows Hello.
Recall uses local AI models built into Windows 11 to screenshot mostly everything you see or do on your computer and then give you the ability to search and retrieve items you’ve seen. An explorable timeline lets you scroll through these snapshots to look back on what you did on a particular day on your PC. Everything in Recall is designed to remain local and private on-device, so no data is used to train Microsoft’s AI models.
Microsoft’s decision to delay Recall comes just after vice chair and president Brad Smith testified before the House Homeland Security Committee today. Smith said that Microsoft is putting security above everything, as part of its Secure Future Initiative (SFI). “It is more important even than the company’s work on artificial intelligence,” says Smith.
Smith also revealed that Microsoft will make security a mandatory part of its bi-annual reviews process for all employees. “With this change, cybersecurity will be considered in every employee’s annual bonus and compensation,” Smith said.
I reported earlier today in Notepad that Recall was originally created before Microsoft’s big SFI overhaul begun. Recall was developed in secret at Microsoft, and it wasn’t even tested publicly with Windows Insiders. Microsoft subsequently identified some of the security issues with Recall and started to develop and test changes to the experience in recent months. It clearly now needs more time to make sure these changes stand up to its promise of putting security above AI and everything else.

Image: Microsoft

Microsoft is planning to launch its new Copilot Plus PCs next week without its controversial Recall feature that screenshots everything you do on these new laptops. The software maker is holding back Recall so it can test it with the Windows Insider program, after originally promising to ship Recall as an opt-in feature with additional security improvements.

“We are adjusting the release model for Recall to leverage the expertise of the Windows Insider community to ensure the experience meets our high standards for quality and security,” says Microsoft in an updated blog post. “When Recall (preview) becomes available in the Windows Insider Program, we will publish a blog post with details on how to get the preview.”

This means that Recall won’t even be available initially to Windows Insiders or anyone who buys a Copilot Plus PC. I wrote in Notepad earlier today that Windows engineers were scrambling to get the security improvements tested and implemented in time for the June 18th launch date of Copilot Plus PCs. Now, Microsoft is essentially admitting here that it needs more time to test Recall’s security improvements.

GIF: Microsoft
The Windows Recall feature.

Microsoft first unveiled the Recall feature as part of its upcoming Copilot Plus PCs last month, but since then, privacy advocates and security experts have been warning that Recall could be a “disaster” for cybersecurity without changes. Microsoft committed to three major updates to Recall last week, including making the AI-powered feature an opt-in experience instead of on by default, encrypting the database, and authenticating through Windows Hello.

Recall uses local AI models built into Windows 11 to screenshot mostly everything you see or do on your computer and then give you the ability to search and retrieve items you’ve seen. An explorable timeline lets you scroll through these snapshots to look back on what you did on a particular day on your PC. Everything in Recall is designed to remain local and private on-device, so no data is used to train Microsoft’s AI models.

Microsoft’s decision to delay Recall comes just after vice chair and president Brad Smith testified before the House Homeland Security Committee today. Smith said that Microsoft is putting security above everything, as part of its Secure Future Initiative (SFI). “It is more important even than the company’s work on artificial intelligence,” says Smith.

Smith also revealed that Microsoft will make security a mandatory part of its bi-annual reviews process for all employees. “With this change, cybersecurity will be considered in every employee’s annual bonus and compensation,” Smith said.

I reported earlier today in Notepad that Recall was originally created before Microsoft’s big SFI overhaul begun. Recall was developed in secret at Microsoft, and it wasn’t even tested publicly with Windows Insiders. Microsoft subsequently identified some of the security issues with Recall and started to develop and test changes to the experience in recent months. It clearly now needs more time to make sure these changes stand up to its promise of putting security above AI and everything else.

Read More 

Photo: NSA

OpenAI has appointed Paul M. Nakasone, a retired general of the US Army and a former head of the National Security Agency (NSA), to its board of directors, the company announced on Thursday.
Nakasone, who was nominated to lead the NSA by former President Donald Trump, directed the agency from 2018 until February of this year. Before Nakasone left the NSA, he wrote an op-ed supporting the renewal of Section 702 of the Foreign Intelligence Surveillance Act, the surveillance program that was ultimately reauthorized by Congress in April.
OpenAI says Nakasone will join its Safety and Security Committee, which was announced in May and is led by CEO Sam Altman, “as a first priority.” Nakasone will “also contribute to OpenAI’s efforts to better understand how AI can be used to strengthen cybersecurity by quickly detecting and responding to cybersecurity threats.”
Recent departures tied to safety at OpenAI include co-founder and chief scientist Ilya Sutskever, who played a key role in Sam Altman’s November firing and eventual un-firing, and Jan Leike, who said on X that “safety culture and processes have taken a backseat to shiny products.”

“Artificial intelligence has the potential to have huge positive impacts on people’s lives, but it can only meet this potential if these innovations are securely built and deployed,“ board chair Bret Taylor said in a statement. “General Nakasone’s unparalleled experience in areas like cybersecurity will help guide OpenAI in achieving its mission of ensuring artificial general intelligence benefits all of humanity.”
OpenAI’s board of directors now includes Nakasone, Altman, Adam D’Angelo, Larry Summers, Bret Taylor, Dr. Sue Desmond-Hellmann, Nicole Seligman, and Fidji Simo. Microsoft’s Dee Templeton also has a non-voting observer seat.

Photo: NSA

OpenAI has appointed Paul M. Nakasone, a retired general of the US Army and a former head of the National Security Agency (NSA), to its board of directors, the company announced on Thursday.

Nakasone, who was nominated to lead the NSA by former President Donald Trump, directed the agency from 2018 until February of this year. Before Nakasone left the NSA, he wrote an op-ed supporting the renewal of Section 702 of the Foreign Intelligence Surveillance Act, the surveillance program that was ultimately reauthorized by Congress in April.

OpenAI says Nakasone will join its Safety and Security Committee, which was announced in May and is led by CEO Sam Altman, “as a first priority.” Nakasone will “also contribute to OpenAI’s efforts to better understand how AI can be used to strengthen cybersecurity by quickly detecting and responding to cybersecurity threats.”

Recent departures tied to safety at OpenAI include co-founder and chief scientist Ilya Sutskever, who played a key role in Sam Altman’s November firing and eventual un-firing, and Jan Leike, who said on X that “safety culture and processes have taken a backseat to shiny products.”

“Artificial intelligence has the potential to have huge positive impacts on people’s lives, but it can only meet this potential if these innovations are securely built and deployed,“ board chair Bret Taylor said in a statement. “General Nakasone’s unparalleled experience in areas like cybersecurity will help guide OpenAI in achieving its mission of ensuring artificial general intelligence benefits all of humanity.”

OpenAI’s board of directors now includes Nakasone, Altman, Adam D’Angelo, Larry Summers, Bret Taylor, Dr. Sue Desmond-Hellmann, Nicole Seligman, and Fidji Simo. Microsoft’s Dee Templeton also has a non-voting observer seat.

Read More 

Image: Cath Virginia / The Verge

Ruthia He, the founder and CEO of the subscription-based “digital health company” Done, and the company’s clinical president, David Brody, “exploited emergency flexibilities during the [covid-19] public health emergency to provide easy access to Adderall and other stimulants that were not for a legitimate medical purpose,” the Department of Justice claims.
The two executives were arrested and indicted for distributing controlled substances, conspiracy to commit healthcare fraud, and other charges. This is the first time the DOJ has charged someone with criminal drug distribution related to their involvement in a telehealth company. The DOJ and DEA began investigating Cerebral, another telehealth company, in 2022. Four months after news of the Cerebral investigation went public, The Wall Street Journal reported that the DEA was also investigating Done.
According to the indictment filed on Thursday in California’s Northern District, He and Brody “conspired to defraud pharmacies and Medicare” by prescribing Adderall and other stimulants to patients who didn’t have ADHD.

“As alleged in the indictment, the defendants provided easy access to Adderall and other stimulants by exploiting telemedicine and spending millions on deceptive advertisements on social media. They generated over $100 million in revenue by arranging for the prescription of over 40 million pills,” Principal Deputy Assistant Attorney General Nicole M. Argentieri, head of the Justice Department’s Criminal Division, said in a statement.
Before the pandemic, the ability to issue prescriptions online was limited by the Ryan Haight Online Pharmacy Consumer Protection Act of 2008, which required medical practitioners to have at least one in-person medical evaluation with a patient before prescribing them any controlled substances. In 2020, the Drug Enforcement Administration implemented temporary exceptions to the Ryan Haight Act — which, charging documents claim, Done executives took advantage of, issuing more than 40 million prescriptions since 2020.

According to charging documents, He, who formerly worked at Meta, made “false and fraudulent representations that Done was a successful business prior to the pandemic” when the company had, in fact, not generated any “material revenue” before 2020.
He and Brody claimed Done could accurately diagnose ADHD with shorter appointments because its screening process weeded out people who were unlikely to have ADHD, the indictment says. But the company also used “deceptive social media advertisements” emphasizing the ease with which Done members could get Adderall prescriptions to intentionally target “drug-seeking patients,” charging documents claim. For a monthly fee, Done offered patients diagnosis, treatment, and refills of ADHD medication, including Adderall.
According to the indictment, He, Brody, and others hired doctors “who they believed were not overly concerned about drug-seeking patients” to issue Adderall prescriptions. Done had a no-follow-ups policy, charging documents claim, and paid doctors on patient load rather than time spent with patients. Instead of requiring subsequent appointments for refills, Done had an auto-refills policy, according to the DOJ. “The purpose, as R. He wrote, was to ‘use the comp structure to dis-encourage follow-up,’” the indictment claims. These allegedly lax prescription policies not only generated hundreds of millions in revenue for Done but also led to overdoses and the death of at least one patient, the DOJ alleges. According to the complaint, one Done member described the company as a “straight up pill mill.”

The complaint also alleges that after media reports claimed Done “made Adderall and other stimulants too easy to obtain” and a grand jury subpoenaed another telehealth company, He and Brody began “altering, destroying, and concealing records and documents” and using encrypted messaging platforms and their personal email accounts instead of their company emails.
If convicted, He and Brody each face up to 20 years in prison. Done did not respond to The Verge’s request for comment.

Image: Cath Virginia / The Verge

Ruthia He, the founder and CEO of the subscription-based “digital health company” Done, and the company’s clinical president, David Brody, “exploited emergency flexibilities during the [covid-19] public health emergency to provide easy access to Adderall and other stimulants that were not for a legitimate medical purpose,” the Department of Justice claims.

The two executives were arrested and indicted for distributing controlled substances, conspiracy to commit healthcare fraud, and other charges. This is the first time the DOJ has charged someone with criminal drug distribution related to their involvement in a telehealth company. The DOJ and DEA began investigating Cerebral, another telehealth company, in 2022. Four months after news of the Cerebral investigation went public, The Wall Street Journal reported that the DEA was also investigating Done.

According to the indictment filed on Thursday in California’s Northern District, He and Brody “conspired to defraud pharmacies and Medicare” by prescribing Adderall and other stimulants to patients who didn’t have ADHD.

“As alleged in the indictment, the defendants provided easy access to Adderall and other stimulants by exploiting telemedicine and spending millions on deceptive advertisements on social media. They generated over $100 million in revenue by arranging for the prescription of over 40 million pills,” Principal Deputy Assistant Attorney General Nicole M. Argentieri, head of the Justice Department’s Criminal Division, said in a statement.

Before the pandemic, the ability to issue prescriptions online was limited by the Ryan Haight Online Pharmacy Consumer Protection Act of 2008, which required medical practitioners to have at least one in-person medical evaluation with a patient before prescribing them any controlled substances. In 2020, the Drug Enforcement Administration implemented temporary exceptions to the Ryan Haight Act — which, charging documents claim, Done executives took advantage of, issuing more than 40 million prescriptions since 2020.

According to charging documents, He, who formerly worked at Meta, made “false and fraudulent representations that Done was a successful business prior to the pandemic” when the company had, in fact, not generated any “material revenue” before 2020.

He and Brody claimed Done could accurately diagnose ADHD with shorter appointments because its screening process weeded out people who were unlikely to have ADHD, the indictment says. But the company also used “deceptive social media advertisements” emphasizing the ease with which Done members could get Adderall prescriptions to intentionally target “drug-seeking patients,” charging documents claim. For a monthly fee, Done offered patients diagnosis, treatment, and refills of ADHD medication, including Adderall.

According to the indictment, He, Brody, and others hired doctors “who they believed were not overly concerned about drug-seeking patients” to issue Adderall prescriptions. Done had a no-follow-ups policy, charging documents claim, and paid doctors on patient load rather than time spent with patients. Instead of requiring subsequent appointments for refills, Done had an auto-refills policy, according to the DOJ. “The purpose, as R. He wrote, was to ‘use the comp structure to dis-encourage follow-up,’” the indictment claims. These allegedly lax prescription policies not only generated hundreds of millions in revenue for Done but also led to overdoses and the death of at least one patient, the DOJ alleges. According to the complaint, one Done member described the company as a “straight up pill mill.”

The complaint also alleges that after media reports claimed Done “made Adderall and other stimulants too easy to obtain” and a grand jury subpoenaed another telehealth company, He and Brody began “altering, destroying, and concealing records and documents” and using encrypted messaging platforms and their personal email accounts instead of their company emails.

If convicted, He and Brody each face up to 20 years in prison. Done did not respond to The Verge’s request for comment.

Read More 

Amazon’s new Alexa has apparently suffered some setbacks. | Illustration by Alex Castro / The Verge

In the voice assistant arms race, the frontrunner may be about to finish last. On the heels of Apple revealing a new “Apple Intelligence”-powered Siri at its WWDC 2024 conference, a new report from Fortune indicates that Amazon’s Alexa — arguably the most capable of the current voice assistants — is struggling with its own generative AI makeover:
… none of the sources Fortune spoke with believe Alexa is close to accomplishing Amazon’s mission of being “the world’s best personal assistant,” let alone Amazon founder Jeff Bezos’ vision of creating a real-life version of the helpful Star Trek computer. Instead, Amazon’s Alexa runs the risk of becoming a digital relic with a cautionary tale— that of a potentially game-changing technology that got stuck playing the wrong game.

The lengthy report (which is paywalled but syndicated in full at Yahoo Finance) draws from interviews with over a dozen former employees, who relayed stories of organizational dysfunction combined with technological challenges that led the company to blow its shot to dominate AI. Fortune reports that Amazon responded to these claims saying the details provided by employees were dated and didn’t reflect the current state of the Alexa LLM.
However, it seems things aren’t going smoothly for the new souped-up Alexa. The more conversational, contextually aware voice assistant the company demoed at its fall hardware event last year still hasn’t rolled out beyond a limited preview. And, according to Fortune’s reporting, while Amazon may eventually launch a better LLM-based Alexa, it won’t be anywhere close to what it could have been.

Image: Amazon

“Alexa, you feeling okay?” A report from Fortune says the new Alexa has some serious growing pains.

Many of the former employees interviewed by Fortune said they left in part because they believed the new Alexa would never be ready or would already be overtaken by competitors if and when it did launch. Its biggest weakness, compared to companies like OpenAI and its headline-grabbing ChatGPT, is that it has to “navigate an existing tech stack and defend an existing feature set,” according to Fortune.
Basically, the old Alexa is getting in the way of the new Alexa. Fortune’s sources say Amazon has not yet figured out how to combine what Alexa can do now with the capabilities it touted for the new Alexa last fall — a better, smarter, more conversational assistant. One employee told Fortune that the message at the company after the demo event was that “we need to basically burn the bridge with the old Alexa AI model and pivot to only working on the new one.”
The message at Amazon was that “we need to basically burn the bridge with the old Alexa AI model and pivot to only working on the new one.”
According to Fortune, Amazon has grappled with getting its Alexa LLM to consistently and effectively make API calls, which is how the current Alexa interacts with your other stuff, such as third-party smart home devices and music services. It’s also struggled to train the LLM to understand natural language, as while it has millions of devices in the wild, its customers have trained themselves to speak in “Alexa language” and don’t interact conversationally with the device.
Another reported hurdle has been Amazon’s decentralized organizational structure, in which the thousands of people working on Alexa are siloed into several teams, causing friction and frustration. Mihail Eric, a research scientist who left the company in 2021, wrote on X (formerly Twitter) that he blames the company’s org chart and insistence research to be tied to a product launch for the failure of his work on Alexa — work he claims that, “if done correctly, could have been the genesis of an Amazon ChatGPT (well before ChatGPT was released).”
For its part, Amazon says it remains committed to its voice assistant’s growth. “Our vision for Alexa remains the same—to build the world’s best personal assistant,” Amazon’s Kristy Schmidt told The Verge in response to Fortune’s article. “Generative AI offers a huge opportunity to make Alexa even better for our customers. We have already integrated generative AI into different components of Alexa, and are working hard on implementation at scale — in the over half a billion ambient, Alexa-enabled devices already in homes around the world — to enable even more proactive, personal, and trusted assistance for our customers. We are excited about what we’re building and look forward to delivering it for our customers.”
Whatever missteps are in its past, it’s clear Amazon is racing to catch up. The former head of devices and services, Dave Limp, left shortly after that fall event. His replacement — Panos Panay, the former chief product officer at Microsoft — has been in place for a little over six months. Fall 2024 is right around the corner. Let’s see if Amazon can deliver on any of its promises.

Amazon’s new Alexa has apparently suffered some setbacks. | Illustration by Alex Castro / The Verge

In the voice assistant arms race, the frontrunner may be about to finish last. On the heels of Apple revealing a new “Apple Intelligence”-powered Siri at its WWDC 2024 conference, a new report from Fortune indicates that Amazon’s Alexa — arguably the most capable of the current voice assistants — is struggling with its own generative AI makeover:

… none of the sources Fortune spoke with believe Alexa is close to accomplishing Amazon’s mission of being “the world’s best personal assistant,” let alone Amazon founder Jeff Bezos’ vision of creating a real-life version of the helpful Star Trek computer. Instead, Amazon’s Alexa runs the risk of becoming a digital relic with a cautionary tale— that of a potentially game-changing technology that got stuck playing the wrong game.

The lengthy report (which is paywalled but syndicated in full at Yahoo Finance) draws from interviews with over a dozen former employees, who relayed stories of organizational dysfunction combined with technological challenges that led the company to blow its shot to dominate AI. Fortune reports that Amazon responded to these claims saying the details provided by employees were dated and didn’t reflect the current state of the Alexa LLM.

However, it seems things aren’t going smoothly for the new souped-up Alexa. The more conversational, contextually aware voice assistant the company demoed at its fall hardware event last year still hasn’t rolled out beyond a limited preview. And, according to Fortune’s reporting, while Amazon may eventually launch a better LLM-based Alexa, it won’t be anywhere close to what it could have been.

Image: Amazon

“Alexa, you feeling okay?” A report from Fortune says the new Alexa has some serious growing pains.

Many of the former employees interviewed by Fortune said they left in part because they believed the new Alexa would never be ready or would already be overtaken by competitors if and when it did launch. Its biggest weakness, compared to companies like OpenAI and its headline-grabbing ChatGPT, is that it has to “navigate an existing tech stack and defend an existing feature set,” according to Fortune.

Basically, the old Alexa is getting in the way of the new Alexa. Fortune’s sources say Amazon has not yet figured out how to combine what Alexa can do now with the capabilities it touted for the new Alexa last fall — a better, smarter, more conversational assistant. One employee told Fortune that the message at the company after the demo event was that “we need to basically burn the bridge with the old Alexa AI model and pivot to only working on the new one.”

According to Fortune, Amazon has grappled with getting its Alexa LLM to consistently and effectively make API calls, which is how the current Alexa interacts with your other stuff, such as third-party smart home devices and music services. It’s also struggled to train the LLM to understand natural language, as while it has millions of devices in the wild, its customers have trained themselves to speak in “Alexa language” and don’t interact conversationally with the device.

Another reported hurdle has been Amazon’s decentralized organizational structure, in which the thousands of people working on Alexa are siloed into several teams, causing friction and frustration. Mihail Eric, a research scientist who left the company in 2021, wrote on X (formerly Twitter) that he blames the company’s org chart and insistence research to be tied to a product launch for the failure of his work on Alexa — work he claims that, “if done correctly, could have been the genesis of an Amazon ChatGPT (well before ChatGPT was released).”

For its part, Amazon says it remains committed to its voice assistant’s growth. “Our vision for Alexa remains the same—to build the world’s best personal assistant,” Amazon’s Kristy Schmidt told The Verge in response to Fortune’s article. “Generative AI offers a huge opportunity to make Alexa even better for our customers. We have already integrated generative AI into different components of Alexa, and are working hard on implementation at scale — in the over half a billion ambient, Alexa-enabled devices already in homes around the world — to enable even more proactive, personal, and trusted assistance for our customers. We are excited about what we’re building and look forward to delivering it for our customers.”

Whatever missteps are in its past, it’s clear Amazon is racing to catch up. The former head of devices and services, Dave Limp, left shortly after that fall event. His replacement — Panos Panay, the former chief product officer at Microsoft — has been in place for a little over six months. Fall 2024 is right around the corner. Let’s see if Amazon can deliver on any of its promises.

Read More 

Apple

Early photos from Apple TV Plus’ upcoming series Sunny left little question about it being a comedy, but the show’s new trailer makes it seem like it’s also going to be a solid (if goofy) whodunit that leaves you guessing.
Set in a semifuturistic modern-day Japan, Sunny follows as Suzie Sakamoto (Rashida Jones) sets out to uncover the truth behind her husband Masa’s (Hidetoshi Nishijima) presumed death. Between Masa’s plane crashing under mysterious circumstances and the disappearance of his body, life already feels weird for Sunny as she settles into her new identity as a widow. But when a colleague of Masa’s shows up at Suzie’s door with a bobble-headed domestic robot named Sunny (Joanna Sotomura) as a bereavement gift, things start to take an even stranger turn.
Grief is definitely going to play a role in Sunny’s more dramatic moments as Suzie and Sunny learn about the secret life Masa kept hidden from his family. But there are enough jokes and weird twists in the trailer to make Sunny look like it’ll be a mystery worth tuning in to when it debuts on July 10th.

Apple

Early photos from Apple TV Plus’ upcoming series Sunny left little question about it being a comedy, but the show’s new trailer makes it seem like it’s also going to be a solid (if goofy) whodunit that leaves you guessing.

Set in a semifuturistic modern-day Japan, Sunny follows as Suzie Sakamoto (Rashida Jones) sets out to uncover the truth behind her husband Masa’s (Hidetoshi Nishijima) presumed death. Between Masa’s plane crashing under mysterious circumstances and the disappearance of his body, life already feels weird for Sunny as she settles into her new identity as a widow. But when a colleague of Masa’s shows up at Suzie’s door with a bobble-headed domestic robot named Sunny (Joanna Sotomura) as a bereavement gift, things start to take an even stranger turn.

Grief is definitely going to play a role in Sunny’s more dramatic moments as Suzie and Sunny learn about the secret life Masa kept hidden from his family. But there are enough jokes and weird twists in the trailer to make Sunny look like it’ll be a mystery worth tuning in to when it debuts on July 10th.

Read More 

Photo by Justin Sullivan / Getty Images

LinkedIn is rolling out new AI-powered features to help users find jobs, tailor their resumes, and even get advice from AI chatbots. But these new features come at a cost, as they’ll be available to Premium subscribers beginning in English worldwide.
One tool lets you search for jobs on LinkedIn by typing a prompt in natural language. In a blog post, LinkedIn chief product officer Tomer Cohen gives the example of “find me a remote marketing job in Detroit that pays at least $110,000.” If the results are good, I could see this being a useful way to seek out potential jobs.

GIF: LinkedIn

LinkedIn can also review your résumé and provide personalized suggestions for improving it for a specific job post. You’ll be able to upload a résumé, get feedback, and “make edits interactively with AI,” LinkedIn’s Rohan Rajiv says in a blog post. There’s a tool to help you build a cover letter “from scratch” with the help of AI, too.
These features add to other Premium-only AI-powered features LinkedIn started testing late last year, which included the ability to summarize posts from your LinkedIn feed.
LinkedIn isn’t just using AI to help you find a job. It’s also beginning to pilot AI personas of a handful of experts that you can have conversations with to learn more about business-y topics. (They seem kind of like LinkedIn’s take on Meta’s celebrity-based chatbots.) “The responses you’ll receive are trained by experts and represent a blend of insights that are personalized to each learner’s unique needs,” Cohen says. The list of instructors includes Alicia Reece, Anil Gupta, Dr. Gemma Leigh Roberts, and Lisa Gates.
The company is also promising that it will improve the platform’s search more broadly with generative AI, though it didn’t share much in the way of specifics. “With our new search capabilities, every search interaction becomes smarter — whether you’re looking to find someone, explore jobs, conduct outreach, or seek knowledge and answers,” Cohen says. “AI is set to revolutionize our search capabilities, enabling you to explore the depth and breadth of any topic directly through LinkedIn search.” Cohen promises that search enhancements will be rolled out “in the coming weeks.”

Photo by Justin Sullivan / Getty Images

LinkedIn is rolling out new AI-powered features to help users find jobs, tailor their resumes, and even get advice from AI chatbots. But these new features come at a cost, as they’ll be available to Premium subscribers beginning in English worldwide.

One tool lets you search for jobs on LinkedIn by typing a prompt in natural language. In a blog post, LinkedIn chief product officer Tomer Cohen gives the example of “find me a remote marketing job in Detroit that pays at least $110,000.” If the results are good, I could see this being a useful way to seek out potential jobs.

GIF: LinkedIn

LinkedIn can also review your résumé and provide personalized suggestions for improving it for a specific job post. You’ll be able to upload a résumé, get feedback, and “make edits interactively with AI,” LinkedIn’s Rohan Rajiv says in a blog post. There’s a tool to help you build a cover letter “from scratch” with the help of AI, too.

These features add to other Premium-only AI-powered features LinkedIn started testing late last year, which included the ability to summarize posts from your LinkedIn feed.

LinkedIn isn’t just using AI to help you find a job. It’s also beginning to pilot AI personas of a handful of experts that you can have conversations with to learn more about business-y topics. (They seem kind of like LinkedIn’s take on Meta’s celebrity-based chatbots.) “The responses you’ll receive are trained by experts and represent a blend of insights that are personalized to each learner’s unique needs,” Cohen says. The list of instructors includes Alicia Reece, Anil Gupta, Dr. Gemma Leigh Roberts, and Lisa Gates.

The company is also promising that it will improve the platform’s search more broadly with generative AI, though it didn’t share much in the way of specifics. “With our new search capabilities, every search interaction becomes smarter — whether you’re looking to find someone, explore jobs, conduct outreach, or seek knowledge and answers,” Cohen says. “AI is set to revolutionize our search capabilities, enabling you to explore the depth and breadth of any topic directly through LinkedIn search.” Cohen promises that search enhancements will be rolled out “in the coming weeks.”

Read More 

Photo by Allison Johnson / The Verge

Google has announced on its Pixel Help Community portal that it’s extending the Pixel 8’s repair program for phones displaying vertical lines and exhibiting flickering. This policy update allows Pixel 8 owners to get this problem fixed under warranty, for up to three years after purchase, if their phone “Exhibits a vertical line running from the bottom of the display to the the top or a display flicker.”
Google isn’t promising that every phone with the issue will qualify, though — Google or an authorized repair partner will need to determine that it qualifies for the extension based on the serial number of your phone.
The Pixel 8 Pro is not included in this repair extension, though there are user claims out there that it can exhibit vertical lines of its own.
Google has a history of users experiencing glitches, bugs, and other oddities in Pixel phones — even dating back to some of the earliest Pixel generations and their Nexus forebears. In fact, shortly after the Pixel 8 and 8 Pro launched in late 2023, users reported ripple-like physical defects appearing in their displays.
Google’s standard hardware warranty still only lasts one year without such an extension, even though the company now promises seven years of software OS and security updates on all Pixel 8 phones, including the cheaper Pixel 8A.

Photo by Allison Johnson / The Verge

Google has announced on its Pixel Help Community portal that it’s extending the Pixel 8’s repair program for phones displaying vertical lines and exhibiting flickering. This policy update allows Pixel 8 owners to get this problem fixed under warranty, for up to three years after purchase, if their phone “Exhibits a vertical line running from the bottom of the display to the the top or a display flicker.”

Google isn’t promising that every phone with the issue will qualify, though — Google or an authorized repair partner will need to determine that it qualifies for the extension based on the serial number of your phone.

The Pixel 8 Pro is not included in this repair extension, though there are user claims out there that it can exhibit vertical lines of its own.

Google has a history of users experiencing glitches, bugs, and other oddities in Pixel phones — even dating back to some of the earliest Pixel generations and their Nexus forebears. In fact, shortly after the Pixel 8 and 8 Pro launched in late 2023, users reported ripple-like physical defects appearing in their displays.

Google’s standard hardware warranty still only lasts one year without such an extension, even though the company now promises seven years of software OS and security updates on all Pixel 8 phones, including the cheaper Pixel 8A.

Read More 

Image: Cath Virginia / The Verge, Getty Images

Tesla shareholders voted today to reapprove Elon Musk’s gigantic pay package, after a Delaware court judge voided it earlier this year, according to preliminary results. Investors also approved a proposal to move the company’s legal home from Delaware to Texas and to reappoint two board members: James Murdoch and Kimbal Musk, Elon’s brother.
The vote was widely seen as a referendum on Musk’s leadership during a time of great upheaval for Tesla, which has seen its stock price slide on news of shrinking sales, softening demand, and declining revenue. The company has laid off at least 10 percent of its workforce, its vehicle lineup is aging and in need of a shake-up, and the success of Musk’s bet on AI and robotics is far from certain.
In spite of all this, Tesla shareholders handed him an enormous victory by approving a compensation package that makes him one of the highest paid chief executives in modern history.
In some ways, the outcome was entirely predictable. Shareholders clearly wanted to reward Musk’s effort to transform Tesla into the most valuable car company in the world, creating enormous wealth for them in the process. And they wanted to refute those who have questioned his leadership choices in recent years, especially as his attention has wandered to other ventures, like the company formerly known as Twitter.
Tesla has an unusually high proportion of retail investors — also known as mom-and-pop investors — compared to large institutional investors. And while several top proxy advisory firms have recommended voting against the proposal, the company’s retail shareholders clearly disagreed.
Even recent negative revelations, like allegations of sexual misconduct by Musk at SpaceX and a sexual harassment lawsuit by fired employees, weren’t enough to affect the outcome. Nor has his political shift to the right and his endorsement of many unhinged conspiracy theories on X. Many investors simply didn’t want to risk the value of their shares plummeting if the pay package was rejected.
But Musk can’t just take the money and run. The Delaware court’s ruling striking down his pay package, on the grounds that Musk had undue influence over its composition, still stands. Today’s vote was part of an unusual strategy by Tesla to correct the flaws in the process that first began in 2018.
“It is not automatically reinstated if the shareholders approve it,” said Samantha Crispin, a partner at Texas-based law firm Baker Botts and chair of the corporate department. “I would imagine it would be influential [and] it certainly would be something that the judge would take into consideration.”
Tesla is banking on the reapproval aiding in its effort to overturn Delaware Chancery Court Judge Kathaleen St. Jude McCormick’s ruling, which means a likely appeal to the state’s Supreme Court. Still, the outcome is uncertain, with Tesla stating in its proxy statement that it “cannot predict with certainty how a vote to ratify Musk’s compensation would be treated under Delaware law in these novel circumstances.”
Still, the vote could help keep Musk’s attention on Tesla — as well as its work on AI and robotics. Earlier this year, Musk threatened to spin out the automaker’s work on artificial intelligence into a separate company if he did not receive 25 percent control of Tesla’s stock. He sold approximately $38 billion worth of Tesla stock in 2021–2022 to finance his acquisition of Twitter and now heads six other companies. With today’s vote, Musk’s stake could potentially rise to 20 percent, from about 13 percent today.
In the run-up to the vote, Tesla used both carrots and sticks to whip the vote. The company dangled an exclusive tour of the Austin factory led by Musk himself as a potential prize for select investors. It also issued an ominous warning that a failure to approve the pay package could spur Musk to abandon Tesla.
The idea of Musk leaving his post at Tesla seems far less likely now that shareholders have confirmed his status as one of the most well-compensated person on the planet. But the saga of his pay package isn’t over yet.

Image: Cath Virginia / The Verge, Getty Images

Tesla shareholders voted today to reapprove Elon Musk’s gigantic pay package, after a Delaware court judge voided it earlier this year, according to preliminary results. Investors also approved a proposal to move the company’s legal home from Delaware to Texas and to reappoint two board members: James Murdoch and Kimbal Musk, Elon’s brother.

The vote was widely seen as a referendum on Musk’s leadership during a time of great upheaval for Tesla, which has seen its stock price slide on news of shrinking sales, softening demand, and declining revenue. The company has laid off at least 10 percent of its workforce, its vehicle lineup is aging and in need of a shake-up, and the success of Musk’s bet on AI and robotics is far from certain.

In spite of all this, Tesla shareholders handed him an enormous victory by approving a compensation package that makes him one of the highest paid chief executives in modern history.

In some ways, the outcome was entirely predictable. Shareholders clearly wanted to reward Musk’s effort to transform Tesla into the most valuable car company in the world, creating enormous wealth for them in the process. And they wanted to refute those who have questioned his leadership choices in recent years, especially as his attention has wandered to other ventures, like the company formerly known as Twitter.

Tesla has an unusually high proportion of retail investors — also known as mom-and-pop investors — compared to large institutional investors. And while several top proxy advisory firms have recommended voting against the proposal, the company’s retail shareholders clearly disagreed.

Even recent negative revelations, like allegations of sexual misconduct by Musk at SpaceX and a sexual harassment lawsuit by fired employees, weren’t enough to affect the outcome. Nor has his political shift to the right and his endorsement of many unhinged conspiracy theories on X. Many investors simply didn’t want to risk the value of their shares plummeting if the pay package was rejected.

But Musk can’t just take the money and run. The Delaware court’s ruling striking down his pay package, on the grounds that Musk had undue influence over its composition, still stands. Today’s vote was part of an unusual strategy by Tesla to correct the flaws in the process that first began in 2018.

“It is not automatically reinstated if the shareholders approve it,” said Samantha Crispin, a partner at Texas-based law firm Baker Botts and chair of the corporate department. “I would imagine it would be influential [and] it certainly would be something that the judge would take into consideration.”

Tesla is banking on the reapproval aiding in its effort to overturn Delaware Chancery Court Judge Kathaleen St. Jude McCormick’s ruling, which means a likely appeal to the state’s Supreme Court. Still, the outcome is uncertain, with Tesla stating in its proxy statement that it “cannot predict with certainty how a vote to ratify Musk’s compensation would be treated under Delaware law in these novel circumstances.”

Still, the vote could help keep Musk’s attention on Tesla — as well as its work on AI and robotics. Earlier this year, Musk threatened to spin out the automaker’s work on artificial intelligence into a separate company if he did not receive 25 percent control of Tesla’s stock. He sold approximately $38 billion worth of Tesla stock in 2021–2022 to finance his acquisition of Twitter and now heads six other companies. With today’s vote, Musk’s stake could potentially rise to 20 percent, from about 13 percent today.

In the run-up to the vote, Tesla used both carrots and sticks to whip the vote. The company dangled an exclusive tour of the Austin factory led by Musk himself as a potential prize for select investors. It also issued an ominous warning that a failure to approve the pay package could spur Musk to abandon Tesla.

The idea of Musk leaving his post at Tesla seems far less likely now that shareholders have confirmed his status as one of the most well-compensated person on the planet. But the saga of his pay package isn’t over yet.

Read More