Google has released its 2023 Data and AI Trends Report, containing a number of points the company expects to become more prevalent in the coming months and years.
In a blog (opens in new tab) discussing the report, Google Cloud Data & Analytics VP Gerrit Kazmaier and Google Cloud Databases VP Andi Gutmans explain how embracing these five trends will help companies stay current and competitive.
The research will also benefit Google in its effort to simplify data integration, increase AI adoption, and meet security requirements.
Cloud and AI trends 2023
In the years leading up to 2026, Google reckons that more than eight in 10 (82%) of organizations will ditch data siloes as they work to implement a more integrated cloud data platform. The company’s cloud arm hopes to improve how its users store, manage, analyze, and govern their data, emphasizing cost optimization.
The research also acknowledges a push toward open-source solutions that prevent lock-ins, which it says is crucial to the fast-paced rollout of AI, speedier development cycles, and boosted ROI.
It’s also predicted that AI will have hit the “tipping point” by 2025, and consumers have already proven how willing to use the technology they are with popular tools like OpenAI’s ChatGPT and Google’s own upcoming Bard. The statistics suggest that by this year, 90% of new enterprise application releases will include embedded AI functionality.
Analytics and insights are also said to receive a significant boost in the near future as businesses look to achieve more effective decision-making, and investments in data and AI have already taken an upward turn over the course of the past 18 months.
Finally, the way companies handle data via cloud storage is in need of significant revision. Numerous regulations and compliance mandates require different types of data to be stored differently, and a blanket approach can leave some things exposed placing companies and their customers at risk. While the precise solution remains to be established, more than three-quarters (77%) of businesses are seeking to improve their ability to classify data and enforce better security and privacy controls.
As we look to the years ahead, it’s clear that how we handle data is going to become even more important, and that artificial intelligence will help streamline many of our existing business operations.
A US government email server was discovered online without a proper password (opens in new tab) to protect its content, essentially leaking sensitive information to anyone who knew where to look. Whether or not anyone really knew where to look – remains to be seen.
The exposed email server was hosted on Microsoft’s Azure government cloud for Department of Defense, allowing it to share sensitive, but still unclassified data.
This service offers servers that are physically disconnected from commercial customers, and was part of an internal mailbox system that held some 3TB of internal military emails, some of which referred to U.S. Special Operations Command (USSOCOM), a military unit running special operations.
Terabytes of data
However it seems that the servier wasn’t protected with a password, so simply knowing the IP address would be enough to access it, and all of the data it hosted.
This hosted data reportedly included sensitive information such as internal military email messages, personal information and health information on certain government employees, and more.
The breach was spotted by security researcher Anurag Sen, who tipped off TechCrunch to the news so that it could alert the US government.
TechCrunch said it had seen some of the data hosted on the server and believes them to be unclassified, “which would be consistent with USSOCOM’s civilian network,” it argues.
The server was first listed as exposed on February 8, but there’s no explanation yet why it happened.
TechCrunch reached out to USSOCOM shortly after, with the server locked down the following day.
Responding to an email inquiry, USSOCOM spokesperson Ken McGraw said that the incident was not the result of a hack: “We can confirm at this point is no one hacked U.S. Special Operations Command’s information systems,” said McGraw.
Game publishing giant Activision has confirmed suffering a data breach late last year.
Despite playing down the seriousness of the importance, many in the media, as well as the underground hacking community, are claiming the breach was serious.
The news broke on Twitter, where cybersecurity researchers vx-underground shared screenshots of stolen data and messages posted by the hackers on Activision’s Slack workspaces. The data was apparently stolen in early December last year, after an Activision employee got their credentials phished, and as a result, the threat actors took internal employee data, and some game data, from company endpoints (opens in new tab).
Downplaying the incident
Activision spokesperson, Joseph Christinat, responded to the news by sharing a statement claiming the company “swiftly” responded to a smishing attack, and “quickly resolved it”. The same statement claims the threat actors did not access any sensitive employee data, game code, or player data.
But vx-underground begs to differ. The files the group shared with TechCrunch shown full employee names, phone numbers, corporate email addresses, and in some cases even postal addresses for their offices. All important information that can be used in identity theft and social engineering attacks.
While the media didn’t outright say it, they hint that Activision tried to hide the data breach. It occurred almost three months ago, yet its employees were also oblivious to the fact. Two current employees spoke to TechCrunch, confirming that this is a problem
“If there is employee’s information involved, they should have disclosed the breach,” one of the employees told the publication.
Activision is one of the world’s biggest game publishers, having published the likes of World of Warcraft, or Call of Duty.
The company is currently working on an acquisition deal with Microsoft, with the latter allegedly agreeing to buy the company out for $68.7 billion. However, regulators in different jurisdictions have opposed the deal, the media said.
The Amazon Kindle has been a real boon for self-publishing authors, but its virtual book store risks being overrun by a particularly prolific new scribe: ChatGPT.
As spotted by Reuters, there are already 200 e-books on Amazon’s Kindle store that list ChatGPT as the author or co-author. But because Amazon doesn’t require that authors disclose whether or not they’ve used AI, that’s likely a huge underestimation of the number of titles that AI tools have either written or co-created.
The ChatGPT-created books are published through Amazon’s Kindle Direct publishing arm, which releases over 1.4 million self-published books (opens in new tab) every year and sells them alongside ones written by big-name authors.
With no mention of AI tools like ChatGPT in Amazon’s Kindle Publishing Guidelines (opens in new tab), or any need for self-publishers to disclose using them, that number could well sky-rocket given the popularity of ‘large language models’ that are particularly skilled at generating text from prompts.
This is great news for ‘authors’ looking to radically boost their output by funneling ideas through the increasingly popular ChatGPT, which is also now available in Bing. In one example highlighted by Reuters, a New York-based salesman with no previous writing experience created a 30-page illustrated children’s book in hours and sold it through Amazon’s self-publishing arm.
But it’s also a potential bombshell for both authors and readers alike, considering Amazon’s virtual shelves could soon be flooded with AI-created books. We’ve asked Amazon for comment on how it plans to tackle this issue and will update this story when we hear back.
For now, Amazon has simply told Reuters that “all books in the store must adhere to our content guidelines, including by complying with intellectual property rights and all other applicable laws.” But it seems clear that it’ll need to go further than that if its Kindle store is to retain a sense of order and help readers distinguish between robo-written books and real authors.
Analysis: Amazon will need to rein in robo-authors
The rapid rise of AI tools like ChatGPT is both a blessing and a curse for self-publishers, depending on how you view the technology. Their text-generating skills are ideal for quickly making first drafts, but authors also risk being drowned in a sea of homogenous look-a-likes.
Given that ‘large language models’ like ChatGPT and Google’s incoming Bard are trained on human-made content, there are also potentially huge copyright issues ahead for works that are predominantly generated using the AI tools.
On the surface, the amount of AI-generated works in Amazon’s Kindle store remains a drop in the ocean of its entire range. But that picture could change quite rapidly, given the sheer number of tutorials on YouTube and Reddit that show how easy it is to create and sell a new book in only a few hours.
Amazon’s aim for Kindle Direct has always been to remove any barriers to entry for fledgling authors and fast-track the creation of new titles. Its own Kindle Direct tips (opens in new tab) page proudly states that “there are no gatekeepers deciding which books are published”.
While that policy is unlikely to change overnight, it may well need to employ some AI detection tools to at least label ChatGPT-generated books for readers. Without some moderation, the best Kindle ereaders could soon see their virtual shelves buckling under the weight of some incredibly prolific new AI authors.
The latest v50 update is finally here for the Oculus Quest 2 and Meta Quest Pro, bringing some very handy features to both VR headsets that should make using them less tedious.
One big update that Quest 2 owners are sure to appreciate is the ability to open 2D apps (such as the Browser) without leaving your current VR experience. This multitasking feature was already available to Meta Quest Pro owners, and makes it much easier to check a walkthrough if you’re stuck in a VR game, or fetch login details from a password manager without needing to take your headset off.
It’s not the biggest upgrade, but it should make using your headset less annoying; although considering that the Quest 2 is less powerful than the Quest Pro, you might find multitasking is a little more sluggish on the older hardware.
The other super-exciting tool packed in with v50 is Direct Touch. This much-needed hand-tracking update makes navigating menus much more intuitive than the existing point-and-pinch system currently employed by the Quest 2 and Pro.
As demonstrated by users (opens in new tab) who’ve tried the feature as part of the v50 beta, when Direct Touch is switched on the Quest’s menus will appear right in front of you, and just as you would on a smartphone you’ll be able to use your hands to tap on buttons and swipe through menus.
Best of all you’ll be able to type on the keyboard just like a real one – as someone who’s used the Quest’s virtual keyboard to write pieces in VR, I can see how this new system will be significantly less tedious and might make working in VR more feasible (though you might still want to pick up a VR-compatible keyboard like the Logitech MX Keys if you’re planning to type a lot).
Hand tracking is one of my favorite VR features, but it’s not quite where I’ve wanted it to be in terms of usability. Direct Touch will do a lot to mitigate my biggest issues with hand tracking, and makes me much less nervous for VR headsets to go without controllers – which the rumored Apple VR headset is expected to do.
Last but not least, v50 will bring a minor improvement to the Meta Quest Pro’s controllers. The Touch Pro handsets are more accurate than Meta’s basic controllers as they’re self-tracking, though this can mean they take a second to work out where they are and start working. The latest update has reduced the amount of time it takes for this initializing step to take, so switching between hand tracking and controllers should take less time.
As with the previous Quest updates from Meta it might take a couple of days, or even weeks, for the new software to hit your system. To check if you’re on v50 (or able to update) you’ll want to head into your headset’s Settings menu, then go to System (the top-left option) and check Software update. Here you’ll see what version you’re running and if the next update is available or not.
Amazon Web Services (AWS) has announced a new partnership with AI experts Hugging Face to help create the ultimate development environment.
The non-exclusive partnership is designed to make it easier to carry out AI workloads using AWS, the biggest cloud computing and cloud hosting provider.
Unlike Google, Microsoft, and OpenAI, which are all focusing their attention on public-facing chatbots, AWS is hoping to work behind the scenes to provide the fundamental tools required to create this sort of technology.
AWS, Hugging Face, and AI
Hugging Face CEO Clem Delangue told Reuters (opens in new tab) that the next generation of the open-source AI model Bloom will be run on Trainium, an AWS-created AI chip.
Chips like Trainium are designed to be more efficient as they are geared toward AI workloads, which should help save developers time and money. In turn, AWS hopes to be able to open up AI workload to even more developers.
“Building, training, and deploying large language and vision models is an expensive and time-consuming process that requires deep expertise in machine learning (ML),” an AWS statement said. “Since the models are very complex and can contain hundreds of billions of parameters, generative AI is largely out of reach for many developers.”
Beyond this, AWS has an extensive history of artificial intelligence, largely centered around its Alexa voice assistant, but also including the large AI model that has been designed to make Amazon search more effective. The company hopes to bring this experience to the table in its new partnership.
Whether other companies like Google and Microsoft should be concerned about this partnership is yet to be seen, however it’s clear that Big Tech is gearing up for an AI-centric year ahead.
Samsung’s in-house voice assistant, Bixby, may not have the user base to rival fellow voice assistants Alexa, Siri and Google Assistant, but the AI-powered helper is getting a suite of much-needed upgrades to further level up the best Samsung phones.
Specifically, Samsung’s Bixby update – which will be rolling out to Samsung Galaxy users throughout February – brings significant improvements to the software’s ability to recognize language.
For instance, English speakers can now answer calls by simply typing a message, which Bixby then converts into audio to be communicated to the caller on their behalf. The feature, called Bixby Text Call, can also be customized to feature AI-generated versions of users’ actual voices, Samsung says.
Bixby Custom Voice Creator allows users to record different sentences that the AI will analyze and turn into custom replies. Samsung’s press release (opens in new tab) notes that the feature is “currently available in Korean” and will soon be rolled out to “other Samsung apps beyond phone calls,” which is slightly confusing wording, so we’ve reached out to the company for clarification on whether it’ll be available in English at launch.
Bixby’s latest update also adds the ability to create a personalized wake-up phrase, though, again, this seems limited to the Korean language for now.
It’s also worth noting that Bixby Text Call – which Samsungs says is available on selected models including the Samsung Galaxy S23, Galaxy S23 Plus, Galaxy S23 Ultra, Galaxy Z Fold 4 and Galaxy Z Flip 4 – requires One UI 5.1 or above for phones operating in English. By the looks of things, Bixby Custom Voice Creator is limited to the Galaxy S23, S23 and S23 Ultra for now.
A much brainier Bixby
In addition to these call-related features, Samsung’s Bixby update also makes the assistant more effective at dealing with follow-up requests.
As the company notes, “Bixby can better understand intent and process follow-up requests by understanding context and associating words previously used in interactions. For example, Bixby users can first launch a workout on Samsung Health and then ask Bixby to play music that best suits that exercise by saying ‘Play music for this workout.’” Neat.
Again, however, it looks like this feature is currently limited to the Korean language, though Samsung notes that “more languages [will] be supported” in the future.
And finally, Bixby’s offline capabilities have been improved. Users can now ask the voice assistant to carry out basic functions – like setting timers, taking screenshots or turning on the flashlight – even when their phone isn’t connected to the web. Apple introduced similar functionality to Siri with iOS 15, so we’re glad to see Samsung bringing its voice assistant in line with the competition here.
Some Wordle games are so difficult that they stick with you long after you’ve completed them. Well, today’s game is one of those – and it nearly did for me.
This was far from my first personal nightmare on Wordle. RUPEE, back in March 2022 was one that I screwed up royally, only escaping with a last-gasp 6/6. HOMER, a couple of months later, caused me similar problems.
The worst of all, for me, was GAMER – a puzzle that I agonized over for more than an hour. I was ultimately left with a choice of two words, GAMER or GAZER, and guessed correctly, heart pounding.
Along the way, I sailed through the likes of CAULK, SWILL, TACIT, WATCH, FEWER and EGRET with 3s, 4s or occasionally 5s – but no real jeopardy. My streak extended onwards to its current value of 415 and I began to wonder if I could complete the remainder of Wordle’s 2,309 puzzles without a failure. That hope almost ended today.
If you’ve not played game #613 yet then go away and do so before you return to read the rest of this – because obviously THERE WILL BE SPOILERS. You might well want a few Wordle clues along the way, too.
All done? Good. Then we can continue.
Wordle #613: A puzzle from hell
The best measure I have of a Wordle puzzle’s difficulty comes from WordleBot, the New York Times’ clever AI tool that analyzes each game.
As well as suggesting the best Wordle starting words to use and explaining where you screwed up (in a rather condescending manner), WordleBot reports an average score for the game across a sample of all those who have played it.
Today’s game achieves an average score of 5.4. Yes, an average score of 5.4. That’s the second highest I’ve ever seen, behind only PARER (6.3, game #454) and level with FOYER (game #304).
For context, most Wordles are in the range 3.5 to 4.5. That makes sense: many people score 3s, 4s or 5s most games, with a few outliers getting 2s, 6s or failures. But an average of 5.4 implies that many, many people will have failed.
In fact, I have a good idea of how many people lost their streaks today, because another thing that WordleBot reports is how many people solved it on each guess.
For today’s game, around 7% didn’t guess it on the final go. Given that’s from a sample size of nearly 400,000 people, we know that at least 32,000 people lost their streaks. I can almost feel the collective frustration.
A look at Twitter adds further evidence, with ‘Wordle 613 X’ trending worldwide and with plenty of people taking to the social media platform to bemoan their luck:
The end of an era. My first ever loss. 412 day streak broken.Wordle 613 X/6⬛🟨⬛⬛🟨⬛🟩⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛⬛🟨⬛🟩🟩🟩🟩⬛🟩🟩🟩🟩 pic.twitter.com/0Zg6b0FyhjFebruary 22, 2023
See more
Wordle 613 X/6⬛🟨⬛⬛🟨⬛🟩⬛🟩🟩⬛🟩⬛🟩🟩⬛🟩⬛🟩🟩⬛🟩⬛🟩🟩⬛🟩⬛🟩🟩all I can say is don’t try the 5 I did…February 21, 2023
See more
Now this is bad luck. When strategy and luck collideWordle 613 X/6⬜🟨🟨⬜⬜🟩🟩⬜⬜⬜🟩🟩⬜🟩🟩🟩🟩⬜🟩🟩🟩🟩⬜🟩🟩🟩🟩⬜🟩🟩February 21, 2023
See more
Why is Wordle #613 so difficult?
The answer to today’s Wordle is RIPER, and even an amateur Wordler can see why that might be a tricky one.
The repeated R is a problem right away, simply because all repeated letters are difficult to find. But the real issue is the fact that it ends in —ER and is therefore one of Wordle’s famous ‘Too Many Answers’ puzzles.
We’ve had plenty of them before, of course – and many have caused similar headaches. FOYER and PARER are two obvious examples but FEWER (average score: 4.9), HOMER (4.7) and RUDER (4.6) weren’t far behind in the difficulty stakes.
The bad news is that there are 141 of these answers in Wordle’s 2,309 games, and we’ve so far only had 31 of them – so expect plenty more over the next few months and years.
It’s actually pretty easy to spot an —ER Wordle, because many of the best start words include both of those letters. Once they’ve turned yellow (or green), you’ll have a good idea of how the game is going to develop.
That doesn’t make it easier to solve, though – as my own game demonstrates.
Despite playing three ‘throwaway’ words that I knew couldn’t be the answer, I still only scraped home on the final guess – there were just too many possible solutions along the way. In fact, WordleBot told me I had 181 after my first guess, and 44 after my second. Even after my fourth guess I still had four to choose from.
It would have been even worse if you were playing on Hard mode, where you have to include letters that are already green or yellow on subsequent guesses. That must have been nearly impossible today.
So, what could I have done differently? Well, WordleBot reckons RIDER would have been a better second guess than LONER, which is a little surprising. Guessing a repeated letter so early seems counterintuitive, after all.
Then again, lots of ER words do have a second R in either first or third position, so ruling that out (or in) as soon as possible could have helped here. Beyond that, I should just have been smarter with my subsequent guesses. For instance, guessing DEVIL wasted a letter, because I’d already used the L. I knew that at the time, but couldn’t see another way to include four other good letters. CUPID would have done a better job, according to WordleBot.
The key thing with an ER word is clearly to plan ahead. You know what trap is coming down the line, so you have to think two or three moves ahead and not just one. Easier said than done, of course.
Still, I got there in the end and kept my streak alive for at least one more day. Here’s hoping tomorrow is a little easier.
VPNs were designed as privacy software to keep you safer online – we all know that. However, in recent years one of the most popular reasons for using the best VPN services has been to unblock content from all over the world.
Whether you’re a Brit abroad looking to watch iPlayer, you want to watch a legal free F1 stream, or even just check out what’s on Netflix in other countries and VPN is the only to do it. What’s more, one of the very best services is currently an absolute steal.
Private Internet Access is the perfect mix of affordability and function. For an incredible $2.19 a month (opens in new tab), you’ll be able to stream tons of global content at the touch of a button – all while staying perfectly private online.
This incredible VPN deal from PIA also includes 2 months free, which means you get 26 months for the price of 24 (opens in new tab). To find out more, simply keep scrolling.
This Private Internet Access VPN deal in full:
What makes this PIA deal so good?
Well, what we love most about this deal is undoubtedly its price. But, PIA delivers some seriously powerful software that lends itself to many different uses.
For starters, its apps are super capable, all open-source, and very easy to use. At the same time, they offer extra customization options so that expert users won’t be disappointed.
Streaming access is a priority of PIA’s, and in our last hands-on review, we saw performance seriously improve. It can now unblock a wide range of Netflix locations and BBC iPlayer, alongside tons of other regional services.
Network coverage isn’t lacking, either. PIA boasts over 10,000 servers across 84 countries – including at least one US VPN server in every state. And, even though it isn’t quite the fastest VPN around, its peaks rose from 450 to 510Mbps with WireGuard the last time we checked. These results should be more than enough for most users.
If reliable security is what you’re after, you’re in good hands with PIA as it offers a good choice of security protocols, a strong kill switch, obfuscation technology, and browser extensions with extra privacy perks.
So, if you’re looking for a streaming VPN that’s also got top-quality security smarts, PIA is a perfect choice – and that bargain price only sweetens the deal. Plus, you can always ask for a refund with its 30-day money-back guarantee if you realize that’s not the right service for you.
Fire-Boltt Phoenix Pro smartwatch is here. Domestic wearable maker Fire-Boltt has launched its latest affordable smartwatch — Phoenix Pro in India. The device comes with Bluetooth calling functionality and offers more than 100 sports modes. The wearable comes with a heart rate sensor and voice assistant support. Price and availability Fire-Boltt Phoenix Pro comes with a price tag of Rs 1,799. Customers can purchase the smartwatch in black, grey and pink colour options The smartwatch can be purchased online from the company website. At this price point, the Fire-Boltt Phoenix Pro will face competition from the recently launched NoiseFit Crew smartwatch. Priced at Rs 1,499, the Noise smartwatch also comes with Bluetooth calling functionality. The smartwatch offers more than 100 sports modes and promises to deliver up to 7 days of battery backup on a single charge.
Fire-Boltt Phoenix Pro smartwatch features Fire-Boltt Phoenix Pro comes with a circular dial and 1.39-inch touchscreen display with 240×240 pixel resolution.. The smartwatch offers Bluetooth calling functionality which lets you make and receive phone calls on the go. The wearable offers more than 120 sports modes. You just need to pick your favorite sport, set goals, and let Phoenix Pro handle everything else by tracking your every move and overall performance. The smartwatch also comes with an upgraded health suite which keeps track of all your important vitals. Along with this, the FIre-Boltt Phoenix Pro also comes equipped with a heart rate sensor and SpO2 monitor. The device is also capable of keeping track of your sleep and other daily activities. Fire-Boltt Phoenix Pro offer more than 100 cloud-based customisable watch faces and promise to offer 7 days of battery backup on a single charge.