Samsung Display looks to be going all in on QD-OLED displays, after it was revealed that the manufacturer is set to increase its production of panels featuring the tech by a massive 50%.
According to think tank UBI Research (via OLED-info), Samsung Display will be increasing the number of QD-OLED panels it produces from 30,000 monthly units currently to 45,000 in 2024.
This is in stark contrast to LG Display, which makes the traditional OLED panel found in practically all the best OLED TVs available today, which was reported to have halted investment in production improvements in 2023. This means that it’s likely that the prices of traditional OLED TVs are set to freeze, opening up an opportunity for the more-expensive QD-OLED tech to drop its price faster, and to start catching up in market share.
Samsung Display also ceased manufacturing LCD panels for TVs in June last year, freeing up production capacity for QD-OLED screens.
Samsung Display currently only offers QD-OLED panels for TVs in three sizes: 55-inch and 65-inch, as seen on the Samsung S95B and class-leading Sony A95K, and now a recently launched 77-inch variety. But the production targets referenced in UBI’s report suggests an even wider selection of screen sizes could be on their way from the company next year.
Analysis: Production bump means more QD-OLEDs… and likely cheaper prices
With recent advancements in black levels, QD-OLED TVs now look like real contenders as an alternative to OLED TVs in terms of performance.
The biggest stumbling block in adoption of the tech currently comes in the form of the hefty asking prices for TVs featuring the panels – based on early prices of the Samsung S95C, it looks like the first TV to feature a 77-inch QD-OLED screen will be priced higher than the equivalent LG G3 flagship TV.
Aside from Samsung, only Sony has made the jump to using QD-OLED panels in its 4K TVs at the time of writing, but this latest news looks likely to open the floodgates for a wider range of brands adopting the tech in the coming years.
Economies of scale mean that the more units Samsung Display produces, the lower the cost of making each unit becomes, which means the TVs themselves can be sold cheaper while still making a profit. And that will mean more companies, such as TCL, get in on the QD-OLED action, which means they’ll be competing with each other for price, pushing things down even further.
With Resident Evil 4’s remake less than a couple of months away, fans have been wondering just how faithful it’ll be to the original game.
In the case of one divisive mechanic, not so much. The Resident Evil 4 remake, launching for PS5 and Xbox Series X|S on March 24, is entirely ditching the quick time events of the original release. In an exclusive with Game Informer (opens in new tab), Capcom confirmed that in their place will be more substantial gameplay elements such as side quests and new enemy types.
These side quests will be scattered throughout the environment as collectible flyers, incentivizing exploration. They offer various objectives like killing certain enemies or solving various puzzles and riddles. What few side quests the original game had, such as shooting a set of blue medallions, are also set to make a return.
(Image credit: Capcom)
New enemy types will be making their debut in Resident Evil 4, too, one of which being a portly villager donning the disembodied head of an ox. And I imagine the village won’t be the only location to house all-new enemy variants. The zealot forces of the castle and militaristic island denizens will hopefully receive new rank and file, too.
The Resident Evil 4 remake isn’t straying too far from its gameplay roots, though, for better or worse. It looks like Ashley’s role in the game will remain much the same. At least in regards to Leon having to protect her from the pursuing hordes.
A strong hint at things to come
(Image credit: Capcom)
Quick-time events were a substantial element of the original Resident Evil 4, and in many cases, weren’t just a cheap substitute for more meaningful gameplay. In fact, their absence in the remake is a staunch clue that certain encounters could play out very differently.
Take the intense knife fight scene with Krauser. In the original game, it’s a cutscene requiring the player to input several button commands under very strict time limits. Failure means having to restart the whole scene from scratch.
Now, that’s quite obviously not going to play out in the same way, given the removal of quick-time events. This could mean the encounter has been reworked into a more involving boss fight, or (and I hope not) been axed entirely. Their absence also hints at other encounters needing to change for the remake.
Take the scenes where Leon has to sprint away from a boulder, for instance. Or needing to dodge the swift tail swipes of the xenomorph-like Verdugo boss. Instances like this are likely to change in the remake, and hopefully for the better.
I’m also really curious to see what kind of variety Capcom brings to Resident Evil 4’s world with the new side quests. There’s really not many to speak of in the original game, besides the aforementioned medallion collection and the shooting galleries. It’ll be interesting to see what kinds of challenges the remake will throw at us.
Customers of Google Fi have been notified via email of “suspicious activity” that may have seen some lower-risk data exposed.
Somewhat ironically, Google Fi promises to be a “private and secure” phone plan that gives its subscribers access to unlimited data, end-to-end call encryption, VPN access, and more, for $50 per month (for one user).
Maybe less secure, though, is its “primary network provider” which informed Google that a “third party system that contains a limited amount of Google Fi customer data” was at risk. The unidentified network provider is likely to be T-Mobile, given that it supports a large portion of the Google Fi network and also experienced a data breach around the same time.
Google Fi data breach
Google told its customers that data including account activation, mobile service plan details, SIM card serial numbers, and account status, are among the key areas said to be at risk.
It stresses that personally identifiable information is excluded from the breach, so anything from a customer’s name, contact details, and date of birth, to financial information and login credentials, should all be safe.
Google’s incident response team says that Google’s systems were not compromised, but that it has worked with the “primary network provider” to implement security measures.
Customers do not need to do anything, and their services continue to work as usual.
Despite this, one Reddit (opens in new tab) user claims to have received an email from Google Fi notifying them that their mobile phone service had been transferred from their SIM card to another for almost two hours.
The author discusses how their primary email account, a financial account, and an authenticator app had all been accessed by a hacker, who may have been able to bypass any SMS-based authentication and gain access to personal accounts.
Google did not immediately respond to TechRadar Pro’s request for comment on the matter.
Samsung Galaxy S23 series: the big camera improvements
A lot of Samsung’s pre-launch hype (like its ‘Epic nights are coming’ promo) has been around the phone’s potential for Moon photography.
But as impressive (and sometimes controversial) as that AI-powered mode is, the bigger quality-of-life upgrade for most people will likely be the S23 Ultra’s improved zoom at more modest focal lengths like 15x zoom.
It seems that the 30x of S23 Ultra is better than S22 Ultra，The clarity is comparable to that of S22 Ultra’s 20x. pic.twitter.com/KlqJVChfsiJanuary 23, 2023
See more
According to a Tweet from the leaker Ice Universe (above), the quality of the S23 Ultra’s 30x zoom will be on a similar level to the S22 Ultra’s 20x zoom.
That’s a big deal, as is the prospect of an improved zoom in the 10x-15x zoom range. In theory, the latter could get close to the quality offered by the folded optics of the S23 Ultra’s 10x zoom lens.
This is the space where Samsung is really opening up a lead on the iPhone. And the prospect of having a slim smartphone with that kind of zoom quality is still an amazing one to me – it wasn’t long ago that Samsung could only get to 10x zoom on a phone by slapping a compact camera on the back of the Galaxy Camera (below). Remember that?
(Image credit: Samsung)
Samsung Galaxy S23 series: what cameras will they have?
As TechRadar’s former Cameras editor, I’m most interested in the photography upgrades we’re going to hear about later – particularly on the Samsung Galaxy S23 Ultra.
These are the likely camera setups we’ll see on each phone according to the leaks.
Samsung Unpacked 2023 is kicking off in an hour – here’s a final reminder of that livestream so you can watch along with us.
When are Samsung’s new phones actually launching?
In terms of when you can physically get your hands on Samsung’s shiny new devices, the latest intelligence suggests that the Galaxy S23, S23 Plus and S23 Ultra will all begin shipping online and in stores from February 17.
It’s not just the chipset being used by Samsung’s hardware that’s been leaked, their other specs might have been revealed early too.
Firstly, in terms of digital storage space, we might see every Samsung Galaxy S23 model will abandon their smallest 128GB variant; that would mean you’d only be able to buy a model with 256GB at a minimum. If these new smallest models adopt prices that are similar to their respective 128GB Galaxy S22 counterparts then we won’t be complaining, but if the change means Samsung’s base models end up being more expensive than before we’ll be disappointed.
The Galaxy S22 Ultra may soon be outclassed by the new Galaxy S23 Ultra flagship (Image credit: Future)
That said, those after a budget option may not be out of luck; another rumor has suggested the regular Galaxy S23 will retain its 128GB version, so we’ll have to see which leak is true.
Another note on storage is that some versions of the phones might be exclusive to the Samsung store. This includes some colors but also the rumored 1TB storage / 12GB RAM model of the Samsung Galaxy S23.
Speaking of RAM; the Samsung Galaxy S23 and Galaxy S23 Plus are expected to only boast 8GB of it, while the Samsung Galaxy S23 Ultra is rumored to offer only 8GB of RAM in its 256GB storage model, but its 512GB and 1TB storage models will deliver 12GB of RAM.
Samsung Galaxy S23 series: performance
If the leaks are to be believed, then Samsung’s Galaxy S23 lineup could be its best-performing smartphones yet.
A big part of that will be down to the Snapdragon 8 Gen 2, a chipset that Samsung hasn’t admitted it will be using but leaks and comments from Qualcomm itself hint that this is indeed what the S23 mobiles will rely on.
If the rumors are true, Apple’s iPhone 14 might be outclassed by the Galaxy S23 (Image credit: Apple)
Leaked Galaxy S23 benchmark scores suggest that the Galaxy S23 (and its Plus and Ultra models) will be the fastest Android phone on the market, with scores that put the device close to the Apple A16 chip used in the iPhone 14 Pro.
These claims are backed up by another source that says all three S23 models will see an increase in processor speed by 36%, an increase in graphics performance by 48%, and an increase in neural processing of 60% compared to Samsung’s S22 phones.
The Snapdragon 8 Gen 2 also claims to give mobile games a bump in performance, with its Adreno GPU promising 25% faster performance and better visuals thanks to its ray-tracing capabilities. We’ll have to wait and see what Samsung announces, but if its new hardware includes the Snapdragon 8 Gen 2 then we could be in for a treat.
Samsung Unpacked 2023: how can you watch?
There’s now just ten minutes to go until the Samsung Galaxy S23 launch. If you’re still wondering how to tune in, you have a few potential options.
Samsung will be streaming Galaxy Unpacked 2023 on the Samsung website (opens in new tab), and the event will also be livestreamed across the company’s usual channels.
That gives us a good idea of where else you’ll be able to watch it. Samsung’s YouTube page (opens in new tab)will definitely host a livestream, for one, and we’ve embedded that below.
The Samsung News site (opens in new tab)is also likely to host the stream, and it’s very possible the video will be viewable elsewhere, too.
Samsung Galaxy S23 series price: how much will they cost?
Given the amount of leaks we’ve digested over the past few weeks, we’re fairly confident about what we expect to see unveiled at this year’s Samsung Galaxy Unpacked event – but just how much the company’s new Galaxy S23 phones will cost is still up for debate.
Based on recent rumors, we suspect Galaxy S23 prices to stay the same as their S22 counterparts in the US (i.e. $799.99, $999.99 and $1,199.99), but the news is less promising for European customers, who could be facing price increases of up to €300 (which would mark the biggest generational increase in Galaxy history!).
(Image credit: WinFuture / Roland Quandt)
One leaker has even suggested that Galaxy S23 prices in Turkey will be equivalent to around $2,660 / £2,145 / AU$3,740 – which would probably be enough to get you two of the company’s new phones in the US and UK. Here’s hoping Samsung doesn’t vary its prices quite so wildly across the globe…
Samsung Galaxy S23 series: the all-important ringtone
Looking for a bit of light relief in the last 15 minutes before Samsung Unpacked kicks off? Check out the series’ new retro ringtone below.
The trademark tune has been around since 2011’s Galaxy S II and this year’s take on it comes courtesy of Yaeji, a Korean-American DJ and producer who straddles the worlds of pop, house music and hip hop. I like its Nintendo-style charm, but as expected it’s divided opinion in the YouTube comments.
Expect to hear this song on a train near you very soon – or perhaps on the music festival circuit, as Yaeji’s booked to play the likes of Coachella later in 2023.
Samsung Galaxy S23 pre-orders: hurry!
If Samsung Galaxy S23 prices do rise, then you’ll want to save every cent you can on the new phones. That’s particularly true if you have your eye on the Galaxy S23 Ultra, because that phone is unlikely to be cheap wherever you are.
The good news is that you can save money on the devices if you act quickly (within the next 20 minutes or so). Galaxy S23 preorders are still live, and the official Samsung site is running a unique pre-registration promo which could save you $50 on one of the phones.
(Image credit: Samsung)
Simply head on over to this page and fill out the ‘register your interest’ form (opens in new tab) to reserve a device. Not only will you secure a firm place in line in doing this, but you’ll also receive $50 of store credit should you commit to a preorder once they go live later today.
But act quickly: the pre-orders will close an hour before the event starts, so you have around 20 minutes to register.
Samsung Galaxy S23 series: price
Regardless of how many leaks there have been around the Galaxy S23 – and seriously, this thing has leaked like a sieve – we don’t know everything about it yet.
For one, we don’t yet know for sure how much it will cost.
(Image credit: Shutterstock / Bazilik production)
The smart money (pun intended) says that in the US at least, all three models will keep the prices from the S22. If true, that would mean prices of $799.99, $999.99 and $1,199.99 respectively, for the base variants.
As Samsung fans will already be well aware, we’re expecting three members of the Galaxy S23 series to arrive in just over an hour: the Galaxy S23, the Galaxy S23 Plus and the Galaxy S23 Ultra.
The leaked image of dummy units you see below (credit: Sonny Dickson) echoes other leaks and rumors around the design direction Samsung is taking with this generation. Each device looks strikingly similar in form to the Galaxy S22 Ultra.
Leaked images of supposed Galaxy S23 family dummy units (Image credit: Sonny Dickson)
The standard and Plus models of the S23 this year appear to shrug off the Contour Cut camera module of the previous two generations of Galaxy S phone, in favor of something a little more sleek and sharp.
TechRadar’s Editor in Chief, Marc McLaren has done a deep-dive on the difference between the three expected models in our Samsung Galaxy S23 vs S23 Plus vs S23 Ultra comparison feature.
What we expect to see at Samsung Unpacked
There have been a lot of rumors flying around, but what exactly are we expecting to see at Unpacked? The main news will be the S23 series, but here’s a broad overview of everything that’s likely en route.
Samsung Galaxy S23: The S23 looks like a relatively minor upgrade on the Samsung Galaxy S22, with the same 6.1-inch FHD+ screen, the same 120Hz refresh rate, and the same rear camera setup. But a new chipset – most likely the Qualcomm Snapdragon 8 Gen 2 – looks a cert, and the design should be brought more in line with the S23 Ultra.
Samsung Galaxy S23 Plus: As with the S23, the Galaxy S23 Plus is likely to be an evolution rather than revolution. Expect a bigger 6.6-inch FHD+ screen and a larger battery than on the vanilla model, but not many other differences.
(Image credit: Samsung)
Samsung Galaxy S23 Ultra: The standout reveal at Galaxy Unpacked should be the Samsung Galaxy S23 Ultra. As well as getting a powerful new chipset it’s tipped to get a whopping 200MP sensor on the rear camera. Elsewhere, a 6.8-inch QHD+ screen, up to 12GB of RAM, up to 1TB of storage, and a 5,000mAh battery should give it the specs to compete with the best phones.
Samsung Galaxy Book 3 family: Rumors suggest that there will be several Galaxy Book 3 models debuting at Unpacked, including the Samsung Galaxy Book 3 Pro, the Galaxy Book 3 Pro 360, and the Galaxy Book 3 Ultra.
One UI 5.1: The only software reveal at the event is likely to be the latest version of Samsung’s One UI. This is unlikely to be a huge release, with bigger changes likely held back for the arrival of Android 14 later this year.
Good afternoon and welcome to our Samsung Galaxy S23 live blog.
We’re now just over two hours from Samsung Galaxy Unpacked, which is due to start at 10am PT / 1pm ET / 6pm GMT today on February 1, or 5am AEDT on February 2 for those tuning in from Australia.
The last-minute leaks have been plentiful, with official imagery, marketing videos and spec sheets raining down to give us a pretty clear idea of what to expect from Samsung’s flagships (on paper, at least).
But while we’ll be serving up a refresher on what we already know about the S23 series, this live blog will also shine a spotlight on the details we don’t know, like their pricing and release dates. We’ll also bring you our expert opinions on features like the S23 camera systems and processors.
Creatives using Linktree as an income source recently got access to paywalls via the Payment lock feature, but the web hosting alternative isn’t stopping there.
They also already have access to a tip jar via Support Me links, but from now on, fans and supporters can thank creatives quickly, easily, and without an account with the “Buy Me a Gift” feature, functioning much like Ko-fi (opens in new tab)’s one-off “buy a coffee” feature.
Plus, building on existing Paypal, Square, and Shopify (opens in new tab) integrations for online store functionality, Linktree users can now add SendOwl content links (made with an account on the service) to a browsable carousel, and display a Bonfire storefront directly on a user’s page to sell merch.
The key difference between the existing tip jar functionality offered by Linktree’s Support Me links and the new Buy Me a Gift feature appears to be that the value of a “gift” can be set manually by a Linktree user. Fans can then select one, two, or even three of these if they choose to support a creator.
Elsewhere, alongside the new additions of embeddable storefronts, from which supporters can easily buy single products or browse all of a user’s products, users can also display single SendOwl links on their Linktree page.
The checkout processes for products bought via SendOwl and Bonfire are hosted by those services, which should mean that they’re reliable and trustworthy.
To get started setting up the new integrations, users should explore the “Make and Collect Money” section of their Linktree account’s admin settings (opens in new tab), or learn more about the integrations from the Help Center pages for SendOwl (opens in new tab) and Bonfire (opens in new tab).
Distributed Denial of Service (DDoS (opens in new tab)) attacks are a growing problem around the world, but organizations in Europe are particularly feeling the sting, new research has found.
A report (opens in new tab) from the Financial Services Information Sharing and Analysis Center (FS-ISAC) and Akamai, claim the number of DDoS attacks against financial firms increased by almost a quarter (22%) year-on-year.
Analyzing the period between November 2021 and November 2022, the researchers found that in Europe alone, the number of DDoS attacks rose by almost three-quarters (73%), making it the region with the highest rise. Of all the DDoS attacks that happened in Europe in this timeframe, roughly half targeted financial institutions.
Crime and hacktivism
DDoS attacks work by sending so many bogus traffic requests to the target website that the server simply can’t handle it all and freezes, crashes, or otherwise becomes unable to provide its services to legitimate customers.
To run a DDoS attack, the criminals need an army of internet-connected devices to send these bogus traffic requests, usually called a botnet. To integrate a device into the botnet, crooks usually infect it with malware. Anything from smart home speakers to routers, can be used as part of the botnet.
Sometimes, threat actors use DDoS attacks as an extortion tactic, promising to stop the assault if a payment is made. In other cases, they use it as a distraction, to run a more destructive attack while the IT team is busy repelling bogus traffic. DDoS is seen as one of the primary weapons in any hacktivist’s arsenal.
A report published in a post (opens in new tab) on the Backblaze blog (via The Register (opens in new tab)) states that the age of a drive was deemed an important factor in predicting failure in correlation with, according to Backblaze’s “cloud storage evangelist” Andy Klein, Backblaze’s aging portfolio of drives.
According to statistics for 230,921 of its drives verified as used for storage purposes, smaller drives (from 4-10TB) failed more often than larger ones (12-16TB), but they were also older.
HDD failure trends
The Backblaze data superficially shows that the smallest Seagate and Toshiba drives in its portfolio are the most prone to failure, with both vendors’ products accounting for 3.64% of all of the company’s drive failures in Q3 2022.
However, Klein did praise the “very respectable” annualized failure rates (AFRs) of its longest standing drives from these vendors (some after almost eight years in service), and went on to cite Seagate drives in particular as suitable for business deployment.
“In general, Seagate drives are less expensive and […] their failure rates are typically not high enough to make them less cost effective over their lifetime.”
Klein believes that, for Backblaze, 2023 will involve replacing its smaller drives with larger ones. So far, there’s no inherent link between the size of a drive and how prone it is to failure, and, in fact, only its 16TB drives saw net decreases in AFR of 0.07%.
As a result, we may have to wait for Backblaze to shift more and more towards larger drives as part of its fleet before any evidence of that sort of correlation begins to surface.
According to Backblaze, its hard drive failure data (opens in new tab) supports the bathtub curve (opens in new tab), whereby a product sees higher rates of failure early in its life cycle, which level out over the course of its life, and increasing again.
However, it has noticed improvements in HDD manufacturing across the board. In a report on the curve from 2021, the company claims that many manufacturers, such as Seagate, have moved to extensive testing of their HDDs before shipping (opens in new tab), decreasing the likelihood of early failures.
It also observed that many drives tend to last longer, failing in their fifth or sixth year far more often than the third or fourth year, as was the case in Backblaze’s previous report on the curve in 2013.
Hackers have been spotted abusing the Microsoft Partner Network feature for Azure AD in an attempt to steal corporate emails and other sensitive data (opens in new tab).
Microsoft and cybersecurity pros Proofpoint worked together (opens in new tab) to combat the threats, explaining how they discovered hackers posing as legitimate companies and successfully getting verified in the Microsoft Cloud Partner Program (MCPP).
Getting verified as a legitimate business allowed the crooks to register verified OAuth apps in Azure AD which were, in reality, malicious and used to steal people’s emails via phishing. To make matters worse, Proofpoint said crooks could have also used this access to steal calendar information, as well.
Running BEC attacks
The threat is particularly worrying as his type of information can be used for cyberespionage, business email compromise attacks, or as a stepping stone towards a more serious form of cybercrime.
Proofpoint seems to have been the first to spot the campaign on December 15, with Microsoft moving in later to disable all fraudulent accounts and apps.
“Microsoft has disabled the threat actor-owned applications and accounts to protect customers and have engaged our Digital Crimes Unit to identify further actions that may be taken with this particular threat actor,” it said in its announcement (opens in new tab).
“We have implemented several additional security measures to improve the MCPP vetting process and decrease the risk of similar fraudulent behavior in the future.”
Microsoft also said it reached out to all affected companies and warned them to thoroughly investigate their environments to make sure they’re safe from compromise.
BleepingComputer says malicious actors have been increasingly using OAuth apps to run “consent phishing” attacks and target business Office 365 and Microsoft 365 data, forcing Microsoft into introducing the “verified” status.
Amazon has revealed that by the end of January 2023, it had set a new corporate record for the most renewable energy purchased by a single company in one year.
The company has long been investing in green energy, and according to a Bloomberg New Energy Finance report, has been the single largest corporate buyer of renewable energy since 2020.
Last year, it added 133 new projects to its portfolio, operating in 11 countries with a combined output of 8.3 gigawatts. That means that last year’s additions alone now account for 41.5% of its 20-gigawatt portfolio.
Amazon renewable energy
The tech giant reckons that these 20 gigawatts of energy, provided by more than 400 projects in 22 countries, are enough to power 15.3 million European homes
The investments in 164 wind farms and 237 rooftop solar projects are designed to support the grids that supply a number of its operations, including its fulfillment centers, retail stores, and AWS data centers.
As well as building its existing portfolios in a number of countries last year, Amazon also entered new countries with its renewable energy ambitions, including Brazil, India, and Indonesia.
All of this, Amazon hopes, will help it reach its target of powering its operations with 100% renewable energy by 2025, which is five years ahead of its initial 2030 target.
The company believes that rapidly scaling renewable energy is one of the most effective ways to combat climate change, and the figures suggest that it’s willing to put its money where its mouth is.
Kyle Harrison, head of sustainability research at Bloomberg New Energy Finance, said that “Amazon’s clean energy portfolio doesn’t just top the corporate charts – it is now among the leading utilities globally as well”. Its readiness to continue pushing despite the lasting effects of the pandemic and subsequent global economic uncertainty should motivate other companies to do the same.
To that point, Amazon isn’t the only company racing to reduce its carbon footprint. Microsoft announced last month that it would be opening a new South Korea project to generate 2.5 gigawatts of solar energy to meet its 2030 renewable-only target.
OpenAI has announced (opens in new tab) a new tool that it says can tell the difference between text written by a human and that of an AI writer – some of the time.
The Microsoft-backed company says the new classifier, as it is called, has been developed to combat the malicious use of AI content generators, such as its very own and very popular ChatGPT, in “running automated misinformation campaigns (opens in new tab), … academic dishonesty, and positioning an AI chatbot as a human.”
So far, it claims that the classifier has a success rate of 26% in identifying AI-generated content, correctly labelling it as being ‘likely AI-written’, and a 9% false positive rate in mislabeling the work of humans as being artificially created.
Spot the difference
OpenAI notes that the classifier performs better the longer the text, and that compared to previous versions, the newer version is “significantly more reliable” at detecting autogenerated text from more recent AI tools.
The classifier is now publicly available, and OpenAI will use the feedback it gets to determine the usefulness of it and to help improve further developments of AI detection tools going forward.
OpenAI is keen to point out that it has its limitations and should not be relied upon as a “primary decision-making tool”, a sentiment shared by most involved in all fields of AI.
As mentioned, the length of the text is important for the classifier’s success, with OpenAI stating that it is “very unreliable” on pieces with less than a thousand characters.
Even longer texts can be incorrectly identified, and human written content can be “incorrectly but confidently labeled as AI-written”. Also, it performs worse on text in written in non-English languages as well as computer code.
Predictable text where the content can only realistically be written one way is also unable to be labelled reliably, such as a list of the first one thousand prime numbers, to give OpenAI’s example.
What’s more, OpenAI points out that AI text can be edited to fool the classifier, and although the classifier can be updated and learn from being tricked like this, interestingly, the company says it is “unclear whether detection has an advantage in the long-term.”
Text that is also very different from that which it has been trained on can cause the classifier issues too, with it “sometimes [being] extremely confident in a wrong prediction.”
On this training data, OpenAI says that it used pairs of written text on the same topic, one AI-produced and the other it believed to be written by a human – some gathered from human responses to prompts used to train InstructGPT, the AI model from the company that is primarily used by researchers and developers.
The development of the classifier comes amid numerous concerns and debates surrounding the use of AI chatbots, such as OpenAI’s own ChatGPT, in academic institutions such as high schools and universities.
OpenAI acknowledges this fact, and has even produced its own set of guidelines for educators (opens in new tab) to understand the uses and limitations of ChatGPT. It hopes its new classifier will not only be of benefit to this institution, but also “journalists, mis/dis-information researchers, and other groups.”
The company wants to engage with educators to hear about their experiences with ChatGPT in the classroom, and they can use this form (opens in new tab) to submit feedback to OpenAI.
AI writing tools have been causing a stir elsewhere too. Tech site CNET recently came under fire for using an AI tool to write articles (opens in new tab)as part of an experiment, but was accused of failing to distinguish theses articles from those written by actual people. Such articles were also found to contain some basic factual errors.
