But can someone learn a language with a chatbot alone?

Research suggests that these tools can help learners build language skills, but there isn’t evidence that strictly using these apps is enough to become fluent in another language.

That said, they can still offer a helping hand. Learning another language involves understanding what you hear and read, as well as knowing how to express yourself verbally and in writing.

I study how chatbots can help someone practice these four skills in different ways.

Chatbots can help build language proficiency

First, people can use the voice features in chatbots such as ChatGPT and Gemini to role-play conversations and practice speaking about different topics. People can practice for a job interview by answering questions about their experience, asking for feedback and trying different ways of expressing an unfamiliar phrase.

Research suggests that practising speaking with chatbots can help language learners become more fluent and accurate. People can repeat a conversation and try again after receiving feedback from the chatbot, giving them time and perhaps the confidence to work on unfamiliar expressions and recurring mistakes.

Speaking another language can be stressful, especially when people worry about making mistakes in front of someone else. Some language learners report finding chatbot conversations less intimidating than interactions with people, and feeling they can experiment without worrying about being judged.

Not being self-conscious may make people more willing to speak and keep practising after they make a mistake.

Chatbots can also support people as they draft a sentence or a paragraph in another language, and offer them feedback and revisions. Research suggests that these tools can help identify grammar errors and improve sentence structure and clarity. Immediate feedback can also encourage learners to revise more often.

But the quality of the feedback that chatbots offer varies. A review of 51 studies on AI feedback in second-language writing found that this kind of feedback could contain errors or misinterpret the writer’s intended meaning. Some studies also reported that learners relied too heavily on AI, taking less control over decisions about their own writing.

Chatbots work best, then, when learners consider which individual suggestions to use and why. For example, they can check whether a suggested change preserves their intended meaning, or ask a chatbot why it was recommended and try using what they learned from the feedback in a new sentence.

People could also ask a voice-enabled chatbot to tell a short story or ask questions, then request that it repeat or rephrase anything they do not understand. In a study of university students in China, language learners practiced listening with ChatGPT and received feedback on their answers outside their regular English classes. Their English listening test scores improved more than those of students who used teacher-provided materials for that additional practice.

Finally, chatbots can help someone work through difficult reading material in a foreign language. Research suggests that reading a simplified version first may help learners grasp the main idea and work out unfamiliar words.

Where chatbots fall short

Chatbots can help people develop some language skills, but they still lack the cultural understanding, lived experience and emotional support that actual human beings bring to a conversation.

For starters, chatbots may reflect cultural biases.

A study of 10 language models compared their answers about the importance of work and religion, among other values, with human survey responses. The chatbots’ answers were generally closer to those of people from the Netherlands, Germany, the United States and Japan, even when asked to respond as someone from India or China.

Some responses also relied on stereotypes. Asked to imagine an ideal job for a Dutch person, Gemini suggested leading bicycle tours past cheese factories, windmills and tulip fields. For language learners, this raises concerns that a chatbot may provide cultural information that is inaccurate, biased or oversimplified.

Privacy matters, too. Companies providing chatbot services may record learners’ text and voice data. Practicing a job interview with real details about an employer, for example, could involve sharing confidential information with a chatbot.

In a survey published in 2025, students who reported more positive experiences with chatbots also worried more about feeling disconnected from other people and lacking emotional support.

In language learning, a 2026 study of 10 English teachers and 10 students at a university in Turkey explored their experiences using AI in the classroom. Some found it useful for teaching and learning, but their accounts also emphasized the importance of connecting with people.

Existing research on using chatbots to learn a language also has limits. Studies in one review tested activities lasting from a single learning session to about four months of practice. Most studies involve adults, and English is by far the most studied language.

Longer studies involving different ages of learners and an array of languages would help establish whether the benefits last and how widely they apply.

Where teachers fit in

As chatbots become more widely used, educators can help students evaluate their responses and make their own decisions about how to use them. For example, a teacher might ask students to write a draft, get feedback from a chatbot and explain why they accepted, adapted or rejected its suggestions.

Teachers also build relationships based on trust and an understanding of students’ needs. They might talk through a difficult task with a discouraged student and offer encouragement to keep going.

Teachers need training to make effective use of chatbots. This includes understanding how they work, recognizing their limitations and learning to design and select activities that meet students’ learning needs.

Chatbots offer new ways to practice a language and build skills, but the evidence does not yet show that they are enough on their own. For now, their clearest value may be offering new opportunities to learn alongside teachers and other speakers.

• Antonio Perez-Nunez is an Associate Professor of Spanish, College of Charleston
• This article first appeared in The Conversation

The post Using an AI chatbot to practice new languages can be helpful, but it doesn’t replace speaking with an actual person appeared first on Stuff South Africa.

Keeping certain chats under your thumb (and only yours)

That’s the biggest selling point of Restricted Chats, but it’s not all they’re capable of. Restricted Chats, judging by the beta test, is only available for individual conversations, and not group or community chats. The setting is off by default across all your chats, so users will need to seek out the toggle on a chat-by-chat basis.

Toggle it, and WhatsApp will apply four restrictions to the chat. First, the app won’t automatically save media sent in said chat right to the device’s gallery. Next, neither user will be able to export the chat. Meta AI is banished to the shadow realm too, and any attempt to use its smarts will fail miserably. These chats stay local, too.

Read More: WhatsApp expands teen safety features with new parental controls

It mimics WhatsApp’s advanced chat privacy settings, though that didn’t support keeping chats off other devices. Any chat with these settings enabled remains accessible on linked devices, defeating the point of privacy. When the feature turns up everywhere, you’ll need to visit the specific chat’s profile to flick the switch.

As for when you’ll be able to do so… WhatsApp isn’t telling. The feature has only just made it to a select few beta tester devices, meaning it’s still a ways off from a public release. It’s also worth mentioning that a beta feature isn’t guaranteed to make it out, though it’s rare for WhatsApp not to follow through.

Source

The post WhatsApp is testing Restricted Chats for an extra layer of security appeared first on Stuff South Africa.

For American and European tech firms, like Microsoft or NVIDIA, the precedent is uncomfortable. Their stock market valuations rest on an assumption of scarcity: capable models being rare, costly and controlled by whoever trained them gives them an edge; a competitive advantage they can capitalise on financially.

Europe, which invented much of the solar industry and then watched its manufacturers disappear, knows what happens when that assumption fails. If a credible AI model can be downloaded for free, the premium on owning one evaporates.

The economics of deliberate loss: a case of déjà vu?

China’s strategy is not unfamiliar; we have watched it unfold twice. In both cases, the science was Western, but the pricing was conquered by a method no private investor would choose: state-owned banks financed the factories regional party committees wanted, and those factories competed by selling below cost. In finance, this is a negative net present value strategy, pouring in more money than you can expect to earn back.

Why would anyone do this? Because the decision-makers are not necessarily trying to make a profit. Andrei Shleifer and Robert Vishny showed that when politicians rather than shareholders steer a company, it serves political ends (employment, prestige, strategic dominance) rather than financial ones.

A private laboratory must earn a return or perish; a state champion need only serve its brief. When the two meet, the player who’s indifferent to losses sets the price. The result is abundance: plentiful supply in the form of alternative comparable models, low prices in the form of more flexible usage policies, and no lasting edge for the innovator, whose profit margins are suppressed. Considering that the two largest AI companies (Anthropic and OpenAI) are preparing for an IPO based on (currently negative) operating profits increasing significantly in the foreseeable future, this is a significant challenge.

Cheapness is also a factor in whether technology is adopted. American laboratories may keep their lead on the elite benchmarks (the standardised tests, such as the GLUE or the MMLU by which AI models are ranked) but those scores do not decide which system ends up embedded in the world’s software. Cost does. A model that is merely good enough and very cheap becomes the default others build upon, and defaults are hard to dislodge once products have formed around them.

But is the same story unfolding for AI?

The analogy deserves scrutiny. Innovation in solar panels and electric vehicles evolves in a discreet manner. New products equipped with better cell chemistry or a denser battery compete based on other qualities, such as cost-effectiveness.

In particular, the evolution of these technological inventions follows electronic engineer Gordon Moore’s 1965 observation (or Moore’s law) that the number of transistors on a chip roughly doubles every two years, so computing power climbs as unit costs fall.

Other technologies, like solar panels or EVs, follow similar curves, albeit at different rates, and see increasing innovation leading to throat-cutting competition and vanishing margins. Anything riding that kind of curve becomes a commodity: so plentiful and interchangeable that nobody can charge much for it.

Does AI ride the same curve?

The models behind artificial intelligence are loosely inspired by the brain: layers of “nodes” whose connections are adjusted until the network produces useful output. But one difference is decisive. An AI model learns once, during training, and thereafter applies what it has learned. Your brain rewires itself with every conversation; today’s AI does not. It might come close with algorithmic trading or with retrieval-augmented-generation, where a trained model adjusts its behaviour according to incoming data, but still, to make it cleverer, you must train it again with more data, more chips, more electricity. Ultimately, more resources.

A trained AI model is therefore a closed system: its abilities are determined the moment training ends. It may surprise us in how it recombines what it knows, but it is unable to pass its current level of performance without being rebuilt or retrained with more data. In a world where AI exhibits notable strategic advantages, like in defence, security, power distribution, etc., the race is on, and the demand for resources follows.

But is a techonology that tries to resemble the cognitive functions of a brain (AI), subject to Moore’s law, like a technology that exhibits no cognitive capabilities (batteries, for example)? Kurt Gödel’s incompleteness theorem offers a useful analogy here. Any closed system is a set of rules and, as such, it cannot prove itself from within. An AI model, in the current state of technology, is restricted in the same way, as it cannot widen its cognitive capabilities using its existing resources. To do so, it needs new resources (training, energy, computation) and, therefore, it is subject to a decay curve, similar to Moore’s law or the ones in solar panels and EVs, destined to meet the same fate.

But can AI deviate from that curve? Should AI models ever become evolutionary in nature (by learning as they process new information in a similar manner to the human brain), then the story changes: such a system compounds its advantage instead of competing for resources. This continuous evolution is the frontier that matters because the competitive advantage of the AI models would be based on which one learns better or faster and not on which one is cheaper.

Why it matters

Financially speaking, if the evolution of AI models remains discrete, its competitive edge will inevitably erode. A (state) champion willing to sell at a loss can push the price of “good-enough” intelligence toward its production cost and margins will fall as they did for solar cells. The implication is uncomfortable and liberating at once: durable returns will come not from owning another interchangeable model but from what is built on top – products, workflows and distribution that rivals cannot copy. As intelligence becomes cheaper, value migrates to whatever stays scarce. China leads the way here; automated drone deliveries (Meituan), driverless taxis (Baidu) or delivery robots (Alibaba) show that value emanates from efficient AI usage and not necessarily from how “smart” a model is.

This logic reframes the race. The industry’s instinct is to build ever-larger AI models burning ever more energy for marginal gains on a benchmark; yet sheer capacity is what abundance renders cheap. The more interesting prize lies with models that learn as they work, because no rival can download a moving target.

So, can AI become a commodity?

On today’s architecture, very probably – and the GLM moment is the warning shot.

The only way to stay ahead of abundance is to change the machine itself: to build intelligence that grows rather than a smarter, static model. Until then, the smart money might follow the example of solar panels: instead of financing the construction of better panels, money goes to how their use can benefit the real economy (energy independence, emissions reduction). Equivalently, instead of pushing for smarter models that might become abundant, profits will arise from smart applications using them, such as traffic control, medical screening, energy grid management, or anything that can create social benefits.

• Iordanis Kalaitzoglou is an Associate Professor in Finance, Audencia
• This article first appeared in The Conversation

The post Made in China AI: new tech, same old story? appeared first on Stuff South Africa.

The company’s decommissioning video, available on YouTube, doesn’t just feature Figure 02 robots jumping into an electric arc furnace. It also has a cameo from the Terminator himself, Arnold Schwarzenegger. The way the robotics outfit tells it, the video was Arnie’s idea.

Humanoid robot father Figure

The decommissioning video starts as a nod to the closing scenes of Terminator 2, where Arnie’s T-800 lowers itself into molten metal to preserve humanity’s future. The whole presentation is fairly sombre, with the Austrian actor himself on hand to give an approving nod… after pressing the button that does the deed. But Figure had more than one robot to dispose of, and those were less careful affairs.

No less complicated, though. F.02 had reached the end of its life cycle, and Figure reckoned that “[d]isassembling each robot would take our technical staff so much time that it would delay the launch of F.04.” The reason for possible disassembly was to protect company information from competitors. Instead, they were simply melted.

The robot lineup was pre-programmed first to perform a last acrobatic stunt as they flung themselves into the furnace below. The results, and the footage, are all spectacular. A humanoid robot cannonballing into a pool of molten metal is impressive. The act is slightly beaten by a backflip into an active volcano, but that’s too hard to film (for now). Figure and Arnie’s plan will have to do.

The post Figure gives its F.02 humanoid robot a Terminator-themed sendoff, complete with a furnace appeared first on Stuff South Africa.

The innovative Pro X2 Superstrike has already been updated; there’s a new version of Logi’s popular wireless gaming headset, a new TKL keyboard with fancy switches aimed at aspiring professionals, a new Yeti microphone for creators, podcasters and streamers, and the cheapest McLaren you’ll ever touch.

Second verse same as the first?

From the outside, Logitech don’t seem to have changed much with its new Pro X3 Superstrike, other than the new colours. To set them apart from the X2 Superstrike, the X3 is available in Black and Magenta colourways.

The important updates live inside. Logitech said it “worked closely with professional athletes” to improve the mouse’s balance and battery life. With its Haptic Inductive Trigger System (HITS) replacing the microswitches you’d usually find in mice, the X2 Superstrike was a little front-heavy.

The Pro X3 Superstrike apparently also features “a 50% increase in power efficiency”. We doubt that will translate into 50% more battery, but even a few more hours would be nice. Whether they’re enough to warrant paying R400 more than the Pro X2 Superstrike remains to be seen. You can judge for yourself, since it’s already available at Computer Mania for R3,900.

Rapid tapping

The Logitech G Pro X2 Rapid is the company’s latest gaming keyboard and packs all the new bells and whistles. TKL means there’s no number pad to get in the way, an 8K polling rate means it sends your keystrokes to your PC before you’ve even thought about pressing them, and the custom magnetic switches allow for adjustable return distance, among other features. And did we mention they’re factory-lubed? How exciting.

Available now at Computer Mania for R4,000.

All about the comms

If you need to get your commands out to the team, the Logitech G Pro X3 Lightspeed and Logitech G Yeti 2 microphone can help with that. The former is Logi’s new wireless gaming headset boasting “premium materials and rigorous quality standards.” It apparently packs a “studio-grade” 48 kHz microphone and offers a 10-band EQ to fine-tune your audio. Those premium materials must include gold or platinum, seeing as it costs R4,900 at Computer Mania.

The Yeti 2 is aimed at creators, podcasters and streamers looking for a standalone mic. It features an “intelligent voice proximity sensor” which is said to track a speaker’s position in 3 dimensions. It’ll also dynamically adjust gain, EQ, and pickup patterns in real time. There’s also AI involved to eliminate background noise, so your teammates have no excuse for not listening to your callouts. Seems like a good deal for R2,900, but we’ll reserve judgement for when one shows up for review.

We have McLaren at home

If you really want to feel a McLaren under your fingertips but don’t have at least R2.2 million to blow on a new car, the Logitech G RS50 McLaren Racing Bundle is one alternative. Sure, you won’t get the same thrill of spinning out in real life, but you also won’t have to deal with enormous insurance premiums and possibly death.

The bundle is a ready-to-race sim setup finished in McLaren Racing’s iconic Papaya aesthetic. It includes an 8Nm RS50 direct-drive wheelbase, a steering wheel that wouldn’t look out of place on a proper track, and load cell pedals. At R16,500, it certainly isn’t cheap. But have you seen what the price of petrol is doing lately?

The post Logitech’s new range of gaming gear lands in South Africa appeared first on Stuff South Africa.

Big. Man.

It beat out the other shortlisted names: Nduna, Orca, Brahman, Chief, and Bakkie, to name a few. All these names tell us that Chery (probably) heavily censored the real list of suggestions, and only went with the first name that wasn’t utterly silly or crass. Personally, we’d much prefer rocketing around in a ‘Bakkie McBakFace’ or something similar.

And rocket you will. This is a plug-in hybrid, equipped with a 2.5L turbo-diesel engine and an electric motor. Pair that with four-wheel drive, and Chery reckons it’ll output 350kW and 800Nm, making it the country’s most powerful bakkie when it goes on sale. Currently, the BYD Shark 6 Performance offers 350kW and 700Nm.

Read More: Arc Ride’s electric Panther bike arrives in South Africa for R22,500

It won’t shy away from towing anything less than 3,500kg and has a groot loading bed, measuring 1,560 x 1,560 x 500 mm. Load it all up, and the bakkie still clears an impressive 230mm off the ground. Chery’s punting this as a proper camping companion, with Vehicle-to-Load functionality allowing the bakkie to power external appliances.

Using just the electric motor, the Grootman should be able to travel 85km WLTP (100km NEDC) before conking out and relying on the diesel engine. At least, that’s the idea. We’ll only get to see how it fares under the boots of real people sometime in 2027, when the Grootman hits SA’s roads. Pricing will be revealed closer to launch.

The post Chery’s Grootman bakkie to bring 350kW of plug-in hybrid power to South Africa in 2027 appeared first on Stuff South Africa.

These are, unsurprisingly, the most competent e-readers Amazon has released in their respective categories. All three of ’em feature flush displays, ditching the recessed look of old. Amazon reckons the basic Kindle is the most “pocket-sized” reader it’s released yet at 6in, though it doesn’t really differ from the 2024 model we reviewed.

Kindle (2026)

The cheapest of the new Kindles was the most deserving of a refresh. Amazon has obliged, delivering a thinner and lighter e-reader that happens to look pretty damn good. This model still has an entirely plastic rear, but an aluminium option is also available — a first for the entry-level reader. You’ll be expected to cough up a hefty premium to get it, though. It and the rest of the Kindle family also introduce edge-to-edge colour, another first.

Possibly our biggest grievance with the 2024 model was the lack of waterproofing, and Amazon hasn’t corrected that here. It has at least promised a 30% speed increase when opening books, on top of the six weeks of battery life. The plastic model only packs in 16GB worth of books, while the aluminium model doubles that figure. Oh, and the power button has been moved to the side — not the top, as the leaks indicated. It’s better than nothing.

The plastic e-reader starts at $150 overseas, converting roughly to R2,500. The premium aluminium model costs $190, or R3,170. That’s a far cry from the $110 starting price we saw when the 2024 model got off the ground.

Kindle Paperwhite (2026)

The 2026 Kindle Paperwhite has also received a redesign, with Amazon claiming it’s thinner and lighter than the last. You’ll notice the power button has also been repositioned to the e-reader’s right-hand side. The 7in flush display is also the largest ever attached to the Paperwhite, and features an oxide-thin-film transistor backplane.

The IPX8 rating has stuck around, meaning it’ll survive a plunge in the bathwater. Adjustable warm light is another promised feature, as is the twelve weeks of battery life off a single charge. Amazon’s selling two models too. The first is a plastic affair with only 16GB capacity, while the aluminium body offers 32GB of storage.

These won’t come cheap. Even the plastic Paperwhite starts at around R3,330 ($200), while the Signature Edition goes up to R4,170 ($250). You’re also paying for auto-adjusting front light and pogo pin dock charging.

Kindle Colorsoft (2026)

Finally, there’s the Kindle Colorsoft. Like the rest of the Kindle family, the Colorsoft now reps a similar redesign. The 7in display is the only one in the line-up to offer colour, which makes comics bearable on Amazon’s handheld e-reader. It’s got an IPX8 rating, adjustable warm light, and up to eight weeks of juice on a single charge.

“Our latest Kindle Colorsoft has a completely redesigned display, using the reverse printing technology we applied across the lineup for edge-to-edge color. The light guide contains micro-deflectors that minimize stray light and improve uniformity, and an ultra-thin optical coating in the display stack improves clarity.”

Once again, Amazon has released two models — one wrapped up in plastic, and another covered in aluminium. The cheaper of the two (plastic) features a $290 (∼R4,830) starting price, with only 16GB of storage. The Signature Edition (aluminium) reaches as high as $320 (∼R5,380) and also doubles the storage. All three models are available for pre-order overseas. There’s no word on a South African launch, but we’re confident they’ll turn up eventually.

The post Amazon refreshes its Kindle line-up for 2026, with three new e-readers on the way appeared first on Stuff South Africa.

The Pro Rugged 13 Extreme does its level best to live up to its title. That is obvious just from looking at it. It weighs around 2.4kg (far less than Durabook’s triple-screened machine) while packing serious elemental and damage resistance. There’s also a fair amount of computing power.

Rugged pull

Starting with the 13 Extreme’s computing performance, buyers can expect at least an Intel Core Ultra 5 335 installed. It’ll spec up to a Core Ultra X7 368H, another of  Intel’s Panther Lake chipsets, while RAM options cover 16GB, 32GB, or 64GB configurations.

Dell’s 13.3in 1,920 x 1,080 touchscreen display goes up to 1,400 nits. That won’t break any visual records, but it should serve for fieldwork in various conditions. The touchscreen is rated for bare or wet fingers as well as gloves, and the panel includes an anti-glare layer. If you’re using one of these inside, you’re doing it wrong. It’ll work… but it’s not supposed to be there.

Why? Because the Pro Rugged 13 Extreme is also rated IP66 (dust and jets of water — no immersion) and has a temperature range of -29°C to 63°C. A drop rating of 1.8 metres should ensure survivability from desks and small rocks (no crevasses, please), while a MIL-STD-461G rating confers some electromagnetic shielding.

Other features include Wi-Fi 7, a Snapdragon X72 5G modem, and a GNSS module as optional extras, a full loadout of ports, plus a configurable I/O bay. The latter supports USB-A, serial, circular connectors, or a physical Ethernet port, depending on what you need.

Dell hasn’t said yet where you can buy one and how much it will cost. The non-Extreme version will set you back at least R85,000 or so. Adjust your wallet accordingly.

The post Dell’s Pro Rugged 13 Extreme laptop comes across as extremely professional (and rugged) appeared first on Stuff South Africa.

Better mechanical hands aren’t much use with no ability behind them. The folks at Boston Dynamics have demoed the new four-fingered appendages (old cartoonists were right, apparently) used both autonomously and tele-operated.

Handy Atlas 

The major upgrade for Atlas is the switch from seven degrees of freedom to thirteen degrees. Human beings, by contrast, have 27 degrees of movement to play with. Even so, the new hand is already producing some impressive results for the dextrous robot.

Boston Dynamics’ demo video (above) shows the robot manipulating a drumstick while working autonomously. It also juggles golf balls in one hand. The company details some of the software training that made these operations possible. Stick a human behind the robot’s control systems, and it gets even more impressive. And disconcerting, since those actions will probably become autonomous eventually.

It’s almost scary seeing Atlas control a hand drill by loading a bit, tightening the chuck, and then drilling a hole in a small piece of wood. Your trip down the uncanny valley will be arrested a little by knowing that the operation was controlled by a human. Given how far the robot has come, however, it won’t be long before it’s able to perform actions like that on its own.

Atlas’ hand upgrade is supposedly to help it prevent damage by improving obstacle avoidance. Just watch how the fingers can splay in the close-up sections in the video above. They’ll also confer new abilities when in the workplace. Our first thought on the upgrade, besides ‘this is terrifying’, was that robots like this would be ideal for setting up a remote base on the moon or Mars. For humans, obviously. What would robots want with their own off-world base?

The post Boston Dynamics’ Atlas gets impressively dextrous new hands appeared first on Stuff South Africa.

That’s cool, Sony, but we haven’t forgotten

Digital Foundry went hands-on with QSSR to test just how worthy it is

Sony says the Pro’s PSSR tech is still the “gold standard.” But without a Pro in your arsenal, QSSR looks like a good deal. QSSR analyses the image pixel by pixel and uses AI to upscale it. The result is a much cleaner image, especially off in the distance. It’s built on two ‘key innovations’, born from its collaboration with AMD: a “streamlined neural network architecture” and a “hand-tuned implementation” to maximise performance.

For the moment, only two PS5 games support the AI tech. Marvel’s Wolverine leads the charge, and Ghost of Yōtei rounds it out. The tech is available to all developers, but it’s on them to implement it. The timing of the QSSR announcement is auspicious, to say the least, with Grand Theft Auto VI right around the corner.

Read More: Sony unveils new PlayStation Pulse and Pulse Edge wireless headsets

“QSSR adds more pixel-level clarity and stability to our environments, bringing out extra detail amidst the busy streets of Madripoor, the dense alleyways of Shinjuku, and many other areas across Logan’s epic journey. It’s a perfect companion to the launch of Marvel’s Wolverine, and it brings a great taste of the PS5 Pro experience to our players on PS5,” said Mike Fitzgerald, head of technology at Insomniac Games.

We’ve yet to test QSSR, but Digital Foundry said the quality boost can be “exceptional,” especially while running Wolverine. Ghost of Yōtei — an already impressive display of fidelity — saw a less obvious visual boost. We’re most looking forward to seeing which older games will get the upgrade in due course.

The post The base PlayStation 5 just got QSSR, a new form of AI upscaling appeared first on Stuff South Africa.