Monthly Archives: May 2022

At the end of its I/O presentation on Wednesday, Google pulled out a one more thing-type surprise. In a short video, Google showed off a pair of augmented reality glasses that have one purpose displaying audible language translations right in front of your eyeballs. In the video, Google product manager Max Spear called the capability of this prototype subtitles for the world, and we see family members communicating for the first time.

Now hold on just a second. Like many people, weve used Google Translate before and largely think of it as a very impressive tool that happens to make a lot of embarrassing misfires. While we might trust it to get us directions to the bus, thats nowhere near the same thing as trusting it to correctly interpret and relay our parents childhood stories. And hasnt Google said its finally breaking down the language barrier before?

In 2017, Google marketed real-time translation as a feature of its original Pixel Buds. Our former colleague Sean OKane described the experience as a laudable idea with a lamentable execution and reported that some of the people he tried it with said it sounded like he was a five-year-old. Thats not quite what Google showed off in its video.

Also, we dont want to brush past the fact that Googles promising that this translation will happen inside a pair of AR glasses. Not to hit at a sore spot, but the reality of augmented reality hasnt really even caught up to Googles concept video from a decade ago. You know, the one that acted as a predecessor to the much-maligned and embarrassing-to-wear Google Glass?

To be fair, Googles AR translation glasses seem much more focused than what Glass was trying to accomplish. From what Google showed, theyre meant to do one thing display translated text not act as an ambient computing experience that could replace a smartphone. But even then, making AR glasses isnt easy. Even a moderate amount of ambient light can make viewing text on see-through screens very difficult. Its challenging enough to read subtitles on a TV with some glare from the sun through a window; now imagine that experience but strapped to your face (and with the added pressure of engaging in a conversation with someone that you cant understand on your own).

But hey, technology moves quickly Google may be able to overcome a hurdle that has stymied its competitors. That wouldnt change the fact that Google Translate is not a magic bullet for cross-language conversation. If youve ever tried having an actual conversation through a translation app, then you probably know that you must speak slowly. And methodically. And clearly. Unless you want to risk a garbled translation. One slip of the tongue, and you might just be done.

People dont converse in a vacuum or like machines do. Just like we code-switch when speaking to voice assistants like Alexa, Siri, or the Google Assistant, we know we have to use much simpler sentences when were dealing with machine translation. And even when we do speak correctly, the translation can still come out awkward and misconstrued. Some of our Verge colleagues fluent in Korean pointed out that Googles own pre-roll countdown for I/O displayed an honorific version of Welcome in Korean that nobody actually uses.

That mildly embarrassing flub pales in comparison to the fact that, according to tweets from Rami Ismail and Sam Ettinger, Google showed over half a dozen backwards, broken, or otherwise incorrect scripts on a slide during its Translate presentation. (Android Police notes that a Google employee has acknowledged the mistake, and that its been corrected in the YouTube version of the keynote.) To be clear, its not that we expect perfection but Googles trying to tell us that its close to cracking real-time translation, and those kinds of mistakes make that seem incredibly unlikely.

Google is trying to solve an immensely complicated problem. Translating words is easy; figuring out grammar is difficult but possible. But language and communication are far more complex than just those two things. As a relatively simple example, Antonios mother speaks three languages (Italian, Spanish, and English). Shell sometimes borrow words from language to language mid-sentence including her regional Italian dialect (which is like a fourth language). That type of thing is relatively easy for a human to parse, but could Googles prototype glasses deal with it? Never mind the messier parts of conversation like unclear references, incomplete thoughts, or innuendo.

Its not that Googles goal isnt admirable. We absolutely want to live in a world where everyone gets to experience what the research participants in the video do, staring with wide-eyed wonderment as they see their loved ones words appear before them. Breaking down language barriers and understanding each other in ways we couldnt before is something the world needs way more of; its just that theres a long way to go before we reach that future. Machine translation is here and has been for a long time. But despite the plethora of languages it can handle, it doesnt speak human yet.

Read more:

Google's AR translation glasses are still just vaporware - The Verge

Theres a good chance youve spent much of the last two-plus years sitting at home, cycling through endless days of virtual meetings staring into your laptops webcam and talking into your built-in mic. This means youve spent much of the last two-plus years appearing to everyone else like a mushy pile of poorly lit pixels, sounding like youre shouting from inside a tin can. Its not your fault: your laptops webcam just sucks. And so does its mic. But Google thinks it can fix them both with AI.

Google announced on Wednesday at its annual I/O developer conference that its Workspace team has been working on a couple of AI-powered ways to improve your virtual meetings. The most impressive is Portrait Restore, which Google says can automatically improve and sharpen your image even over a bad connection or through a bad camera. Portrait Lighting, similarly, gives you a set of AI-based controls over how youre lit. You cant move the window off to your left, Google seems to be saying, but you can make Google Meet look as if you had one to your right as well. And when it comes to sound, Googles rolling out a de-reverberation tool meant to minimize the echoes that come from talking into your laptop from a boxy home office.

A lot of the underlying tech here comes from the AI and machine learning work Google has done with its Pixel phones. Those have substantially better hardware to work with than your average laptop webcam, but Prasad Setty, the companys VP of digital work experience, said the principle is the same. We want to make sure that the underlying software does the same thing, that we are able to use it across a wide range of hardware devices, he said.

As hybrid and remote work have grown, the Google Workspace team has spent the last couple of years thinking about how to make work a little easier, Setty said. We want technology to be an enabler, Setty said in an interview. We want it to be helpful, we want it to be intuitive, and we want it to solve real problems. That led the Workspace team to think more about collaboration hence the meeting tools but also about how to make asynchronous work more palatable.

Googles planning to roll out a new tool that generates automatic summaries of Spaces activity, so you can log on in the morning and catch up without having to read hundreds of messages. Its also launching an automated transcription service for Meet meetings, with plans to eventually also summarize those.

We want to be able to help people handle this information overload, Setty said, and use AI to do so. He also said Googles thinking a lot about collaboration equity and representation equity, trying to help keep everyone on an equal playing field no matter where they are, what tech theyre using, or how theyre working. One trick for Google, Setty acknowledged, is in helping people without getting too involved or making employees feel like theyre being watched by either Google or their employer. The way we think about it, he said, is we want to empower users first and foremost. And then give them like the choice of like how they expose that information to their teams and so on.

After all this time stuck at home, its nice to have a few tools to make your setup work a little better, especially ones that dont require new apps or gear. But as people go back to the office, Google has an even bigger meeting challenge ahead: solving the problem of the hybrid meeting, with some people in a room and others on a screen. Thats going to take a lot more than good lighting and de-reverb.

Excerpt from:

Google is adding tools to make Meet video calls look better - The Verge

Google is getting back in the tablet game. The companys internal hardware division plans to release an Android-powered tablet in 2023, senior vice president of devices and services Rick Osterloh announced on Wednesday at its I/O developer conference. Osterloh was light on details, except to say itll run on the same Tensor processor inside Googles latest Pixel phones and that he imagines it as a consumer-focused gadget focused on entertainment and consumption rather than work. (The Verges Dan Seifert, who briefly saw a picture during a product briefing ahead of I/O, immediately said it looks like an old Samsung tablet.) But Osterlohs overall message is clear: Google cares about Android tablets. For real this time.

The announcement is a huge about-face from Googles recent history. In 2019, when Osterloh himself said Google was getting out of the tablet business. Hey, its true, he tweeted in response to rumors that Google was shutting down its existing tablet products, Googles HARDWARE team will be solely focused on building laptops moving forward, though he again said the software teams still care about supporting tablets. That announcement seemed to chill the market, as if signaling that Google was never going to get serious about tablets. Since then, few companies outside of Samsung have continued to make Android slates.

So why the change of heart? We see it as a critical part of how people are interacting with media and computers at home, especially, Osterloh said in an interview ahead of I/O. Google seems to think of Chrome OS as mostly a tool for work and school, while Android is the consumer product. (Osterloh did say Google plans to make more Pixelbooks, by the way, but he wouldnt say when.) And the pandemic made it clearer than ever to Google that tablets have a unique place in users lives as entertainment, gaming, and general consumption devices. (Youd think a decade-plus of Apples iPad success would have made that clear already, but alas.) And certainly, we think we want to design something thats a perfect companion for Pixel with a larger form factor, he said.

One way to understand the Pixel lineup in general is less as a product organization designed to sell lots of devices and more as a showpiece of Googles intentions. Google isnt just building a flagship tablet it hopes lots of people will buy; its building a flagship tablet so that Android developers in and out of Google will believe Google actually cares about these devices and will have a reason to care about how their apps look on a larger screen. Google cant build a great tablet if it cant start that flywheel. Solving its tablet problem will take more than great hardware and even more than dedicated focus from the entire team at Google. It requires the whole ecosystem to decide its worth giving a crap when Google has given them a decade of reasons not to.

Luckily, Google knows that better than anyone. In the early days of Android, back when Googles in-house hardware projects were called Nexus rather than Pixel, the Nexus 7 and Nexus 10 were some of the best tablet hardware on the market. But Android tablets never caught on, in large part because Android was never well-suited to tablets and Google never seemed to care. Apps mostly treated tablets like phones with humongous screens rather than a device category unto themselves, so they showed up big and misfit and generally didnt work right.

Internally, at least, that appears to be changing. The Android team has been investing a ton in the space, Osterloh said. Starting with Android 12L and continuing with some new features in Android 13, like an updated taskbar with an app drawer that makes it easier to use two apps side by side, the OS is at long last starting to look optimized for larger screens. The Pixel Tablet (or whatever its called) will be a showcase for that work. We wanted to make sure people knew about it so that they knew not only is this a big thing for the Android team, but its a big thing for our team, too. And we intend to be in this category, starting then and going forward forever.

Trystan Upstill, a VP of engineering on the Android team, said that his team has also been working with third-party developers on adapting their apps to the larger displays. We have TikTok, Zoom, and Facebook building out new apps this year for tablet, he said in an interview, and many more coming as well. Googles also updating 20 of its own apps Google Maps, Google Messages, YouTube Music, and more for tablets over the next few weeks. Thats a strong sign of progress, but the fact that those updates are only coming now says a lot about Googles history in the space.

Googles focus on tablets may also have something to do with the state of smartphones. As foldable phones become more popular, theyre also becoming one of the most compelling reasons to pick an Android device over an iPhone. Except, all too often, Android doesnt work great on the larger foldable screens, either. The distinction, of course, is they have two screens, Upstill said. And the layout changes make a big difference. But what were doing for tablets is translating directly to improvements for foldables as well.

And if youre rooting for Android tablets to be great, you should be rooting for foldables to be successful. Because its not hard to imagine a world in which Google really tries to make Android tablets work again only to discover that Apples iPad and Microsofts Surface are so far ahead that its going to be impossible to catch up and decide to just move on to something newer and shinier. Foldables could keep Googles focus on big screens much longer, though, if they are indeed the next big thing in smartphones.

Osterloh seems to be serious about bringing the Pixel shine to the Android tablet world. But if Googles going to pull off a successful launch next year, its going to need help. And fast.

Read more:

Google's Pixel team is working on an Android tablet for 2023 - The Verge

While its against the law in the US to install stalkerware on an adults phone without consent, marketing such apps is legal. Although many companies display disclaimers on their websites stating that their software is intended for legal purposes only, there have been a handful of convictions for installing spyware on unknowing adults devices.

Last September, in the first order of its kind, the Federal Trade Commission banned a company called Support King, which operated under the name SpyFone, from the surveillance business for illegally harvesting and sharing peoples private information and failing to implement basic security measures. The FTC said it will be aggressive about seeking surveillance bans when companies and their executives egregiously invade our privacy.

While many stalkerware apps are sold as parental monitoring tools for keeping an eye on children, they provide the same capabilities as services that are more blatant about being designed to spy on spouses, says David Ruiz, senior privacy advocate at the security group Malwarebytes. Theres a whole family of applications out there that straight up says they will, quote unquote, solve your problem of a cheating spouse. Which is not just ludicrousits dangerous.

Technology-facilitated abuse is a rapidly growing problem. Around 1.5 million Americans are stalked through some form of technology every year, according to the Stalking Prevention Awareness and Resource Center, while the UK domestic violence charity Refuge reported a 97% increase in the number of abuse cases requiring specialist tech support between April 2020 and May 2021.

The charitys tech abuse team said it works with countless survivors whose abusers installed stalkerware on their phones in a bid to intimidate, harass, and manipulate them.

To hear that these apps are being marketed directly to perpetrators is extremely concerning, says Emma Pickering, tech abuse lead at Refuge. Tech companies must act swiftly to remove ads which enable perpetrators to access tools to read their partners messages or track locations without their knowledge or consent.

We must recognize that cyberstalking is dangerous and threatening behavior in the same way stalking on the street is.

Read more:

Google is failing to enforce its own ban on ads for stalkerware - MIT Technology Review

Faisal Masud knows what it takes to make a multibillion-dollar company thrive.

The 49-year-old spent more than two decades working his way up the executive ranks at Amazon, Google, eBay and Staples. Now, he's trying to implement lessons from those successful firms as the CEO of Fabric, a Seattle-based e-commerce start-up that launched in 2016 and is valued at $1.5 billion. (The company stylizes its name as "fabric," to avoid confusion with online insurance company Fabric Technologies.)

Masud was recruited as CEO in 2020 by co-founder Ryan Bartley, with whom he'd worked closely at Staples. Upon stepping into the role, Masud says, he worked to take what he'd learned from his previous employers to create a culture at Fabric that champions empathy, efficiency and most of all success.

"Culturally, we've built a company that's sort of a hybrid of all the companies I've worked at," Masud tells CNBC Make It. "We're able to take the best pieces out of the places I've had experiences and apply those."

Fabric makes e-commerce software that essentially competes with platforms like Shopify and Salesforce, though Masud notes that Fabric is tailored specifically for mid-size and large businesses. That pitch is apparently music to the ears of investors, who have plowed north of $293 million into the company.

Part of Masud's success at Fabric has been a long time coming. He says he learned how a work culture could impact a company's bottom line from Amazon, where he spent nearly a decade following the dot-com bubble.

At Fabric, Masud preaches "ownership," one of Jeff Bezos' famous 14 key leadership principles: Having a team of employees or a single team leader "own" an idea or project can make the decision-making process more streamlined and effective. Masud says he's worked in environments without ownership, and things got messy.

"When something failed, it was nobody's fault. When something was successful, everybody got to celebrate," he says. "That's not how start-ups work. Someone has to make a decision."

After Masud left Amazon, he spent years at Google and eBay, picking up a new workplace skill that stood out in direct contrast to Amazon's culture: empathy.

"Being very empathetic towards your employees was also very important, because that's something Google does really well, and so does eBay, whereas Amazon was not that great at that," says Masud, who was a senior director of shipping at eBay until 2012 and COO of Alphabet's Project Wing drone delivery division from 2018 to 2020.

Today, the "No. 1 value" Masud tries to instill at Fabric is born from those lessons in workplace empathy: "Seek to understand before being understood." Ensuring a "built in" sense of empathy across the company, he says, ultimately leads to better end results for the bottom line.

"Listening attentively, versus just because you have to, brings about a thought process that's different to just executing quickly," he says. "Our culture is sort of embedded in the fact that we're good listeners. But we also use data and facts to make our decisions at the end of the day."

Working with high-profile start-up founders like Bezos and Google's Sergey Brin also taught Masud the value of long-term thinking and staying focused on only the core areas where your company excels.

"These founders are always looking extremely long term," Masud says. "They're always looking at: 'OK, what's the ultimate goal of where the company's got to go?' And then find a way to avoid the distractions along the way."

Masud brought along several high-level employees who also previously worked at Amazon to help instill some of those lessons at Fabric. He refers to them as the "Amazon mafia," and their experience with their former employer could come in handy: Amazon recently launched a "Buy With Prime" service that rivals platforms like Shopify and Fabric.

Bringing in ex-Amazon employees doesn't necessarily mean Masud is trying to recreate Amazon's culture. "I don't think every Amazonian is the right fit for Fabric," he says, adding that he instead tried to "hand select" specific people who agreed with his vision for Fabric's culture.

Together, Masud says, the lessons he learned at his previous workplaces prepared him surprisingly well to be a CEO: "As one of our investors put it: 'You worked for the last two decades to build Fabric. You just didn't know it.'"

Sign up now: Get smarter about your money and career with our weekly newsletter

Don't miss:

How an ex-Blue Origin intern got a $500,000 check from Mark Cuban to build a major SpaceX rival

Girls Who Code 'made coding cool' but toxic tech culture means 'there's still such a long way to go'

Read the original:

This 49-year-old CEO used lessons from Amazon and Google to build a $1.5 billion start-up - CNBC

Tech giant Google is preparing to purge nearly 900,000 apps, which have been abandoned or not been updated, from the Play Store.

According to Android Authority, the Google Play Store could see the number of available apps drop by nearly a third.

Google and Apple have both unveiled measures to deal with abandoned apps or apps that have not been updated in two years. In Google's case, that amounts to 869,000 apps, while Apple has some 650,000.

According to CNET, Google is preparing to hide those apps, making it impossible for users to download them until the developers update them.

The main reason both companies are taking these measures is to protect their users' security.

Older apps do not take advantage of changes in Android and iOS, new APIs, or new development methods that bring enhanced protection. As a result, older apps can have security flaws that newer apps don't have, the report said.

--IANS

vc/vd

(Only the headline and picture of this report may have been reworked by the Business Standard staff; the rest of the content is auto-generated from a syndicated feed.)

Business Standard has always strived hard to provide up-to-date information and commentary on developments that are of interest to you and have wider political and economic implications for the country and the world. Your encouragement and constant feedback on how to improve our offering have only made our resolve and commitment to these ideals stronger. Even during these difficult times arising out of Covid-19, we continue to remain committed to keeping you informed and updated with credible news, authoritative views and incisive commentary on topical issues of relevance.We, however, have a request.

As we battle the economic impact of the pandemic, we need your support even more, so that we can continue to offer you more quality content. Our subscription model has seen an encouraging response from many of you, who have subscribed to our online content. More subscription to our online content can only help us achieve the goals of offering you even better and more relevant content. We believe in free, fair and credible journalism. Your support through more subscriptions can help us practise the journalism to which we are committed.

Support quality journalism and subscribe to Business Standard.

Digital Editor

Read more:

Google to remove nearly 900,000 abandoned apps from Play Store - Business Standard

Google has outlined its vision for this years major Android update, which looks set to continue many of the customization and privacy initiatives that the search giant introduced with last years Android 12. Its customizable Material You color schemes will now be available as preset themes and are also expanding to cover third-party apps icons and the media player. There are also new security features, including a dedicated Privacy and Security menu.

The direction isnt likely to come as much of a surprise to anyone whos kept up with Android 13s early betas. But todays announcements, made to coincide with the search giants annual Google I/O developer conference, see the company lay out its overarching vision for this years major Android update. The search giant is releasing Android 13s second public beta today to coincide with the announcements.

After last years Material You customizable themes feature, Android can already match its color scheme to that of your phones wallpaper. This year, media controls are also receiving a similar Material You-style overhaul and will be able to extract colors from the album art of the music being played. Another new feature for those that dont want or need their phones theme to exactly match their wallpaper is a series of optional preset color schemes to choose from.

Material You theming options are also coming to third-party app icons, which appeared in Android 13s first developer preview in February. This was a bit of a missing part for us in the last release, explains Googles vice president of product management, Sameer Samat. It felt like everything in the system UI got that nice Material You treatment except the icons. For us, it always felt like unfinished business. The new app icon-theming options will come to Pixel devices first and only work with supported apps.

Google Messages RCS support is also set to get a big improvement later this year with the beta launch of end-to-end encryption for group chats, a feature that is currently only available in one-to-one RCS chats in Google Messages. The search giant says the standard, which aims to be a successor to the now-ancient SMS and MMS protocols, is now available to over 500 million Google Messages users worldwide.

As weve seen from its betas, Android 13 is also placing more restrictions on the personal data and phone features apps can use by default. Soon, apps will have to ask for permission to even send notifications in the first place, and theres also a new photo picker that lets you restrict the photos and videos an app can access, rather than granting permission to see your whole library. New permissions will also limit apps to accessing either Photos & videos or Music & audio files, rather than all filetypes.

A new Security and Privacy settings page is being added later this year to collect all of your critical data privacy information in one place. Its designed to encourage Android users to address any security issues that might crop up.

Away from Android phones themselves, Google is also emphasizing the work its doing on interconnectivity with other devices. It plans to add fast-pairing support for the incoming Matter smart home standard this fall to make it quick and easy to use an Android phone to add supported smart home devices to your network. Support for the new power-efficient Bluetooth LE Audio standard is also on the way in Android 13.

A final feature worth mentioning: Android 13 will let users set system languages on a per-app basis, a feature that Samat says is helpful for multilingual users who rely on different languages in different situations. If youre using a social media app, you might use one language. But if youre banking, you might use another language, he explains.

After the chaotic rollout of Android 12, its perhaps reassuring to see that Googles focus this year is on refining rather than revolutionizing Android. Theres no massive change of direction here, just a steady series of tweaks and improvements to Androids existing initiatives.

Follow this link:

Google's vision for Android 13 is to offer a little more of everything - The Verge

For Google, a company that built its reputation on organizing the worlds information, the latest sales pitch to users is that it will try to do more with less of it.

At its I/O 2022 developer conference on May 11th, the tech giant announced a range of privacy measures that it says will help users retain more control over how their data is used by Google applications and displayed to the world through search.

One new change introduced at the conference is the My Ad Center interface: a hub that will let users customize the types of ads they see by selecting from a range of topics they are interested in or opt to see fewer ads on a given topic.

Google says that My Ad Center will help to give users control not just over how their data is used but also over how this affects their experience of the web.

In another announcement unveiled at the conference, Google said that users would be able to request that personal information such as email or address details be removed from search results through a new tool that will be accessible from a users Google profile page.

Perhaps expectedly for a conference geared toward developers, some of Googles most significant privacy announcements involved changing approaches to software engineering. The safety and security segment of the event, led by Jen Fitzpatrick, Googles SVP for core systems and experiences, emphasized the concept of protected computing: a set of technologies that Google says represent a transformed approach to where and how data is processed.

In summary, protected computing means that more data will be processed on devices (e.g., Android phones) without being sent to Googles cloud servers. And when user information is sent to Googles servers, more of it will be anonymized through techniques like the use of differential privacy and edge computing.

Fitzpatrick said that the changes were about justifying the trust that users put in Google to keep them secure.

Protecting your privacy requires us to be rigorous in building products that are private by design, she said.

The safety and security presentation included an acknowledgment that users expectations of privacy are changing and that the company has a need to recognize and adapt to them. Its notable that Google is increasingly trying to prove to users that it can keep at least some of their data out of the hands of the advertisers that bring in the vast majority of the companys revenue.

And under the guiding statement, secure by default, private by design, Google is also pushing to boost user safety across its products by implementing additional security measures out of the box.

Security announcements made at the I/O event included a number of measures meant to increase user protections across a range of Google products. For one, a new account safety status icon will show a warning on a users profile across all Google apps when any security issues are identified and direct the user toward recommended actions to correct the issue.

And the company will expand two-step verification for accounts by sending an Is this you? notification to phones when a user tries to sign in to a Google account elsewhere on the web.

Phishing protection will also be coming to the Google Workspace suite, with the Docs, Sheets, and Slides applications soon to display warning notifications about malicious links in documents.

Overall, Googles safety announcements suggest a company that wants to be seen as centering users security concerns. At an I/O event full of new and creative uses of user data, its heartening to see that, on the face of things, privacy was by no means forgotten.

Follow this link:

Google pitches for user trust with expanded privacy controls - The Verge

It is Google I/O 2022 this week, among many other things, and we were hoping for an architectural deep dive on the TPUv4 matrix math engines that Google hinted about at last years I/O event. But, alas, no such luck. But the search engine and advertising giant, which also happens to be one of the biggest AI innovators on the planet because of the ginormous amount of data it needs to make use of, did give out some more information about the TPUv4 processors and systems that use them.

Google also said that it was installing eight pods of the TPUv4 systems in its Mayes County, Oklahoma datacenter, which is kissing 9 exaflops of aggregate compute capacity, for use by its Google Cloud arm so researchers and enterprises would have access to the same kind and capacity of compute that Google has to do its own internal AI development and production.

Google has operated datacenters in Mayes County, which is northeast of Tulsa, since 2007 and has invested $4.4 billion in facilities there since that time. It is located in the geographic center of the United States well a little south and west of it and that makes it useful because of the relatively short latencies to a lot of the country. And now, by definition, Mayes County has one of the largest assemblages of iron to drive AI workloads on the planet. (If the eight TPUv4 pods were networked together and work could span all at simultaneously, we could possibly say the largest unequivocally. . . . Google surely did, as you will see in the quote below.)

During his keynote address, Sundar Pichai, who is chief executive officer of the Google and also of its parent company, Alphabet, mentioned in passing that the TPUv4 pods were in preview on its cloud.

All of the advances that we have shared today are possible only because of continued innovation in our infrastructure, Pichai said talking about some pretty interesting natural language and immersive data search engine enhancements it has made that feed into all kinds of applications. Recently, we announced plans to invest $9.5 billion in datacenters and offices across the US. One of our state of the art datacenters is in Mayes County, Oklahoma and I am excited to announce that there, we are launching the worlds largest publicly available machine learning hub for all of our Google Cloud customers. This machine leaning hub has eight Cloud TPU v4 pods, custom built on the same networking infrastructure that powers Googles largest neural models. The provide nearly 9 exaflops computing power in aggregate, bringing our customers unprecedented ability to run complex models and workloads. We hope this will fuel innovation in across fields, from medicine to logistics to sustainability and more.

Pichai added that this AI hub based on the TPUv4 pods already has 90 percent of its power coming from sustainable, carbon free sources. (He did not say how much was wind, solar, or hydro.)

Before we get into the speeds and feeds of the TPUv4 chips and pods, it is probably worth it to point out that, for all we know, Google already has TPUv5 pods in its internal-facing datacenters, and it might have a considerably larger collection of TPUs to drive its own models and augment its own applications with AI algorithms and routines. That would be the old way that Google did things: Talk about generation N of something while it was selling generation N-1 and had already moved on to generation N+1 for its internal workloads.

This doesnt seem to be the case. In a blog post written by Sachin Gupta, vice president and general manager of infrastructure at Google Cloud, and Max Sapozhnikov, product manager for the Cloud TPUs, when the TPUv4 systems were built last year, Google gave early access to them to researchers at Cohere, LG AI Research, Meta AI, and Salesforce Research, and moreover, they added that the TPUv4 systems were used to create the Pathways Language Model (PaLM) that underpins the natural language processing and speech recognition innovations that were the core of todays keynote. Specifically, PaLM was developed and tested across two TPUv4 pods, which each have 4,096 of the TPUv4 matrix math engines.

If the shiniest new models Google has are being developed on TPUv4s, then it probably does not have a fleet of TPUv5s hidden in a datacenter somewhere. Although we will add, it would be neat if TPUv5 machinery was hidden, 26.7 miles southwest from our office, in the Lenoir datacenter, shown here from our window:

The stripe of gray way down mountain, below the birch leaves, is the Google datacenter. If you squint and look off in the distance real hard, the Apple datacenter in Maiden is off to the left and considerably further down the line.

Enough of that. Lets talk some feeds and speeds. Here, finally, are some capacities that compare the TPUv4 to the TPUv3:

Last year, when Pichai was hinting about the TPUv4, we guessed that Google was moving to 7 nanometer processes for this generation of TPU, but given that very low power consumption, it is looking like it is probably etched using 5 nanometer processes. (We assumed Google was trying to keep the power envelope constant, and it clearly wanted to reduce it.) We also guessed that it was doubling up the core count, moving from two cores on the TPUv3 to four cores on the TPUv4, something that Google has not confirmed or denied.

Doubling the performance while doubling the cores would get the TPUv4 to 246 teraflops per chip, and moving from 16 nanometers to 7 nanometers would allow that doubling within roughly the same power envelope with about the same clock speed. Moving to 5 nanometers allows the chip to be smaller and run a little bit faster while at the same time dropping the power consumed and having a smaller chip with potentially a higher yield as 5 nanometer processes mature. That the average power consumed went down by 22.7 percent, and that jibes with an 11.8 percent increase in clock speed considering the two-and-change process node jumps from TPUv3 to TPUv4.

There are some very interesting things in that table and in the statements that Google is making in this blog.

Aside from the 2X cores and slight clock speed increase engendered by the chip making process for the TPUv4, it is interesting that Google kept the memory capacity at 32 GB and didnt move to the HBM3 memory that Nvidia is using with the Hopper GH100 GPU accelerators. Nvidia is obsessed about memory bandwidth on the devices and, by extension with its NVLink and NVSwitch, memory bandwidth within nodes and now across nodes with a maximum of 256 devices in a single image.

Google is not as worried about memory atomics (as far as we know) on the proprietary TPU interconnect, device memory bandwidth or device memory capacity. The TPUv4 has the same 32 GB of capacity as the TPUv3, it uses the same HBM2 memory, and it has only a 33 percent increase in speed to just under 1.2 TB/sec. What Google is interested in is bandwidth on the TPU pod interconnect, which is shifting to a 3D torus design that tightly couples 64 TPUv4 chips with wraparound connections something that was not possible with the 2D torus interconnect used with the TPUv3 pods. The increasing dimension of the torus interconnect allows for more TPUs to be pulled into a tighter subnet for collective operations. (Which begs the question, why not a 4D, or 5D, or 6D torus then?)

The TPUv4 pod has 4X the number of TPU chips, at 4,096, and has twice as many TPU cores, which we estimate to be 16,384; we believe that Google has kept the number of MXU matrix math units at two per core, but that is just a hunch. Google could keep the TPU core counts the same and double up the MXU units and get to the same raw performance; the difference would be how much front end scalar/vector processing needs to be done across those MXUs. In any event, at the 16-bit BrainFloat (BF16) floating point format that the Google Brain unit created, the TPUv4 pod delivers 1.1 exaflops, compared to a mere 126 petaflops at BF16. That is a factor of 8.7X more raw compute, balanced against a factor of 3.3X increase in all-to-all reduction bandwidth across the pod and a 3.75X increase in bi-section bandwidth across the TPUv4 interconnect across the pod.

This sentence in the blog intrigued us: Each Cloud TPU v4 chip has ~2.2x more peak FLOPs than Cloud TPU v3, for ~1.4x more peak FLOPs per dollar. If you do the math on that statement, that means the price of the TPU rental on Google Cloud has gone up by 60 percent with the TPUv4, but it does 2.2X the work. This pricing and performance leaps are absolutely consistent with the kind of price/performance improvement that Google expects from the switch ASICs it buys for its datacenters, which generally offer 2X the bandwidth for 1.3X to 1.5X the cost. The TPUv4 is a bit pricier, but it has better networking to run larger models, and that has a cost, too.

The TPUv4 pods can run in VMs on the Google Cloud that range in size from as low as four chips to thousands of chips, and we presume that means across an entire pod.

Go here to read the rest:

Google Stands Up Exascale TPUv4 Pods On The Cloud - The Next Platform

Following last years makeover of the Google TV app, which added features like movie and TV recommendations, critics scores and more, the company today announced at its Google I/O developer conference that, later this year, users will be able to cast TV shows and movies directly from their Android phone or tablet to their TV. The company additionally offered updated metrics related to its Android OS growing footprint and introduced new Android OS developer tools.

Google has been working to overhaul its connected TV experience and the companion app since 2020, when it first introduced the new Google TV interface for its Chromecast streaming devices. It also applied the Google TV name to the app that was previously known as the Google Play Movies & TV app. Earlier this year, Google also removed the Movies & TV section from the Play Store, noting the Google TV app would be the centralized place to buy, rent and watch movies and TV shows on mobile devices.

The company didnt provide additional details about its plans to add support for casting to the app, saying those would come closer to launch. But in a photo Google provided, it showed at least one of the supported apps is NBCUs Peacock.

Image Credits: Google

As part of its announcements, the company also offered a few updates on the Android TV ecosystem, noting there are now 110 million monthly active devices on Android TV OS, including Google TV. Thats up from the 80 million monthly active devices figure Google announced at last years Google I/O event.

It also said the Android TV OS now offers over 10,000 apps.

Of course, the Android TV OS is not a direct equivalent to something like Roku or Amazon Fire TV, as its not only used on the companys first-party devices, like Chromecast. Instead, Googles strategy is to license its platform to partners, including TV OEMs and pay-TV operators worldwide. Currently, there are over 300 partners using the platform, including seven out of the 10 largest TV OEMs, Google said.

But Googles 110 million active devices figure cant be directly compared with the metrics shared by rivals Roku and Amazon.

Android TV OS figures are actually calculated by counting the number of devices that were actively used in a month which means a user with multiple devices could have those devices counted separately, but a family with multiple people watching on one device would be counted once.

Roku and Amazon define monthly active users as accounts that have been active during the month. That means, even if that account streams on several different devices during the time period, it would only be counted once. If Roku or Amazon were to calculate active devices as Google is doing, their numbers would be higher.

In December 2020, Amazon said Fire TV had topped 50 million monthly active users a number still referenced today on its Amazon Ads website.This January, Amazon also noted it had sold more than 150 million Fire TV devices to bring its metrics more in line with Googles device claims. Roku, meanwhile, said during it had ended 2021 with 60.1 million active accounts. It updated that figure to 61.3 million accounts in Q1 2022. It doesnt count active devices.

Image Credits: Google

Google also highlighted several developer features and tools on Android 13 related to its Android TV efforts. The tools are focused on performance and quality, accessibilty and multitasking, Google said, and include:

The picture-in-picture API is particularly interesting as it could introduce more co-watching functionality across the Android TV ecosystem, following Apples introduction of SharePlay, which allows co-watching over FaceTime across platforms, including Apple TV.

More here:

Google TV app to add casting as Android TV ecosystem grows to 110M monthly active devices - TechCrunch