Here's what went wrong.

I can't believe College Football is $60 off.

WIRED tested the popular AI video generator from OpenAI and found that it amplifies sexist stereotypes and ableist tropes, perpetuating the same biases already present in AI image tools.

MacworldAt the heart of every Apple device is an Apple processor. Apple has been using its own chips in its iPhones and iPads for more than a decade, while Apple silicon in the Mac is already in its fourth generation.Whats remarkable about Apple silicon is its performance and power efficiency. But all chips arent created equally. Understanding the performance differences between each chip will help with your buying decisions, especially when youre deciding between iPhone 16 or MacBook models. Knowing how each chip performs gives you a better idea of what products to buy and whether or not its worth your money to step up to a higher model.Lets take a look at how the new processors compare with the rest of the processors in the iPhone, iPad, and Mac lineup and see how each performs and what that means to you. For the sake of consistency, weve usedGeekbench 6benchmarks. Heres every chip and how the benchmarks compare with each other.Update March 23, 2025: Added benchmarks for the M3 Ultra chip; added the new iPhone 16e, iPad, iPad Air, Mac Studio, and MacBook Air.Every current processor comparedResults are scores. Higher scores/longer bars are faster. Chips in this chart are currently available in Apple devices.Before we get into the individual processors, lets let the chips fall where they may. In the above chart, weve only included chips that are in Apples current product lineups for the sake of keeping the chart manageable. The Mac section below includes all of the chips, from the M1 to the current chip. If youre looking for scores of chips that are no longer being used in Apples active iPhone or iPad lineups (such as the A12 Bionic), check out the Geekbench Browser.Its a somewhat predictable chart, with the fastest Mac chips at the top, followed by a mix of iPads and iPhones. But there are still some fascinating results: owners of the iPad Pro can say their tablet is about as fast as a MacBook Air and that wouldnt be much of a reach. And the difference between the $599 iPhone 16e and the $999 iPhone 16 Pro isnt as huge as their price difference indicates.If youre not seeing all the bar chart labels, it may be because your browser font is set larger than the default, or your browser is zoomed in. Youll need to set the font size and browser view to the default to see all the chart labels. Read about how Apples M series processors compare to Intel in our Mac processor guide.iPhone processorsResults are scores. Higher scores/longer bars are faster. Chips in this chart are currently available in Apple devices.Lets look at the specifications of the iPhones currently in Apples lineup so we can understand the differences between them.ProcessorPerformance coresEfficiency coresGraphics coresNeural EngineMemoryThermal Design PowerDevicesA18 Pro2 at 4.04GHz4 at 2.2GHz616-core8GB10WiPhone 16 ProiPhone 16 Pro MaxA182 at 4.04GHz4 at 2.2GHz516-core8GB9WiPhone 16iPhone 16 PlusA182 at 4.04GHz4 at 2.2GHz416-core8GB9WiPhone 16eA16 Bionic2 at 3.46GHz4 at 2.02GHz516-core8GB6WiPhone 15Specifications of chips used in current Apple iPhones.Not surprisingly, the A18 Pro in the iPhone 16 Pro is the fastest. The difference between the A18 Pro and the A18 in the iPhone 16 is that the A18 has one fewer GPU core. The iPhone 163 has two fewer GPU cores than the iPhone 16 Pro.Apple iPhone 16 ProRead our reviewPrice When Reviewed:1199Best Prices Today:1089 at Alternate | 1119 at Computeruniverse | 1119 at cyberportiPad processorsResults are scores. Higher scores/longer bars are faster. Chart includes chips in discontinued Apple devices.The staggered release of Apples iPad lineup creates an odd-looking performance order of CPU and its device.ProcessorPerformance coresEfficiency coresGraphics coresNeural EngineMemoryTransistorsThermal Design PowerDevicesM44 at 4.4GHz6 at 2.851016-core16GB28 billion20W13 & 11 iPad ProM43 at 4.4GHz6 at 2.851016-core8GB28 billion20W13 & 11 iPad ProM34 at 3.49GHz4 at 2.06GHz916-core8GB20 billion15W13 & 11 iPad AirA17 Pro2 at 3.78GHz4 at 2.11GHz516-core8GB19 billion8WiPad miniA162 at 3.46GHz3 at 2.02GHz416-core6GB11.8 billion6WiPad (11th gen)Specifications of chips used in current Apple iPads.The M4-equipped iPad Pros are the fastest models, and the gap between them and the iPad and iPad mini is significant. Furthermore, the M4 is 1.5 times faster than the M2 that it replaced in the previous iPad Pros.The 11th-gen iPad that was released in the spring of 2025 has an A16, an upgrade from the A14 Bionic in the previous model.Apple iPad mini (A17 Pro)Read our reviewPrice When Reviewed:599 EuroBest Prices Today:554 at notebooksbilliger.de | 569 at Alternate | 569 at ComputeruniverseMac processorsResults are scores. Higher scores/longer bars are faster. Chart includes chips in discontinued Apple devices.With Apples M-series of chips for the Mac, the companys release schedule involves the base version in the MacBook Air, 13-inch MacBook Pro, Mac mini, and iMac. Apple then modifies it to create higher-end versions.The latest M-Series chip is the M4, which was released with the new iMac, Mac mini, and the MacBook Pro in the fall of 2024. The M4 Pro and Max were also released in the MacBook Pro, replacing the M3 Pro and Max in those laptops. The M3 Ultra is now in the Mac Studio but Mac Pro still uses the M2 Ultra. The MacBook Air uses the M4 chip.ProcessorPerformance coresEfficiency coresGraphics coresNeural EngineBase memoryTransistorsThermal Design PowerDeviceM3 Ultra24 at 4.52GHz8 at 2.59GHz8032-core96GB184 billion140WMac StudioM3 Ultra20 at 4.52GHz8 at 2.59GHz6032-core96GB184 billion140WMac StudioM4 Max12 at 4.52GHz4 at 2.59GHz4016-core48GB70W14 & 16 MacBook ProM4 Max10 at 4.52GHz4 at 2.59GHz3216-core36GB62W14 & 16 MacBook ProM4 Pro10 at 4.52GHz4 at 2.59GHz2016-core24GB46W14 & 16 MacBook Pro. Mac miniM4 Pro8 at 4.52GHz4 at 2.59GHz1616-core24GB38W14 MacBook Pro, Mac miniM2 Ultra16 at 3.49GHz8 at 2.4GHz7632-core64GB134 billon80WM2 Ultra16 at 3.49GHz8 at 2.4GHz6032-core64GB134 billon80WM44 at 4.41GHz6 at 2.59GHz1016-core16GB28 billion22WiMac, 14 MacBook ProM44 at 4.41GHz4 at 2.59GHz816-core16GB28 billion20WiMacSpecifications of chips used in current Apple Macs.The M4 Max is a beast of a chip, blazing in both CPU and GPU performance but its not the fastest. The M2 Ultra is in the Mac Pro, which has PCIe expansion slots. If you dont need such slots, you can opt for an M3 Ultra Mac Studio. The M3 Ultra is Apples fastest Mac.Apple 14-inch MacBook Pro (M4)Read our reviewBest Prices Today:1678 at notebooksbilliger.de | 1678 at OTTO | 1719 at ComputeruniverseThe chip that started it all, the good ol M1, may seem slow compared to Apples more current chipsbut thats not to undermine Apples original Mac processor. Remember, the M1 blows past the Intel processors it replaced, resulting in a significant price/performance value.

Large language models (LLMs) are rapidly transforming into autonomous agents capable of performing complex tasks that require reasoning, decision-making, and adaptability. These agents are deployed in web navigation, personal assistance, and software development. To act effectively in real-world settings, these agents must handle multi-turn interactions that span several steps or decision points. This introduces the need for training methods beyond simple response generation and instead focuses on optimizing the entire trajectory of interactions. Reinforcement learning (RL) has emerged as a compelling approach to train such agents by refining their decision-making based on long-term rewards.Despite their potential, LLM-based agents struggle with multi-turn decision-making. A major challenge lies in assigning proper credit to actions taken at earlier stages of interaction, which influence later outcomes. Traditional training methods rely on next-token prediction or imitate high-probability actions, which do not account for long-term dependencies or cumulative goals. As a result, these methods fail to address the high variance and inefficiency of long-horizon tasks, particularly in collaborative scenarios where understanding human intent and reasoning across multiple steps is critical.Various reinforcement learning techniques have been adapted to fine-tune LLMs, especially from single-turn human feedback scenarios. Tools like PPO, RAFT, and DPO have been explored but exhibit significant limitations when applied to sequential interactions. These methods often fail at effective credit assignment across turns, making them less effective for multi-turn decision-making tasks. Benchmarks used to evaluate such tools lack the diversity and complexity required to assess performance in collaborative, real-world settings robustly. Value-based learning approaches are another alternative, but their need for custom heads and large amounts of task-specific fine-tuning data limit their generalization capabilities.FAIR at Meta and UC Berkeley researchers proposed a new reinforcement learning method called SWEET-RL (Step-WisE Evaluation from Training-time Information). They also introduced a benchmark known as CollaborativeAgentBench or ColBench. This benchmark is central to the study, providing over 10,000 training tasks and over 1,000 test cases across two domains: backend programming and frontend design. ColBench simulates real collaboration between an AI agent and a human partner, where agents must ask questions, refine their understanding, and provide iterative solutions. For programming, agents are required to write functions in Python by asking for clarifications to refine missing specifications. In front-end tasks, agents must generate HTML code that matches a visual target through feedback-based corrections. Each task is designed to stretch the reasoning ability of the agent and mimic real-world constraints like limited interactions, capped at 10 turns per session.SWEET-RL is built around an asymmetric actor-critic structure. The critic has access to additional information during training, such as the correct solution, which is not visible to the actor. This information allows the critic to evaluate each decision made by the agent with a much finer resolution. Instead of training a value function that estimates overall reward, SWEET-RL directly models an advantage function at each turn, using the Bradley-Terry optimization objective. The advantage function determines how much better or worse a particular action is compared to alternatives, helping the agent learn precise behaviors. For example, if an action aligns better with the human partners expectation, it receives a higher advantage score. This method simplifies credit assignment and aligns better with the pre-training architecture of LLMs, which rely on token-level prediction.SWEET-RL achieved a 6% absolute improvement over other multi-turn reinforcement learning methods across both programming and design tasks. On backend programming tasks, it passed 48.0% of tests and achieved a success rate of 34.4%, compared to 28.2% for Multi-Turn DPO and 22.4% for zero-shot performance. On frontend design tasks, it reached a cosine similarity score of 76.9% and a win rate of 40.4%, improving from 38.6% with DPO and 33.8% with fine-tuning. Even when evaluated against top proprietary models like GPT-4o and O1-Mini, SWEET-RL closed the performance gap significantly, enabling the open-source Llama-3.1-8B model to match or exceed GPT-4os frontend win rate of 40.4%.This research demonstrates that effective training of interactive agents hinges on precise, turn-by-turn feedback rather than generalized value estimations or broad supervision. SWEET-RL significantly improves credit assignment by leveraging training-time information and an architecture-aligned optimization approach. It enhances generalization, reduces training variance, and shows strong scalability, achieving better results with increased data. The algorithm also remains effective when applied to off-policy datasets, underlining its practicality in real-world scenarios with imperfect data. The research team created a meaningful evaluation framework by introducing ColBench as a benchmark tailored for realistic, multi-turn tasks. This combination with SWEET-RL provides a strong foundation for developing agents that can reason, adapt, and collaborate effectively over extended interactions.Several key takeaways from this research include:SWEET-RL improved backend programming success rates from 28.2% (DPO) to 34.4% and frontend win rates from 38.6% to 40.4%.It allowed Llama-3.1-8B to match the performance of GPT-4o, reducing dependency on proprietary models.The critic uses training-time information (e.g., correct solutions) that is invisible to the actor, creating an asymmetric training setup.Tasks in ColBench are capped at 10 rounds per session and include over 10,000 procedurally generated training examples.ColBench measures outcomes using unit test pass rates (for code) and cosine similarity (for web design), providing reliable evaluation.SWEET-RL directly learns a turn-wise advantage function, improving credit assignment without needing an intermediate value function.The model scales effectively with more data and performs well even on off-policy datasets from weaker models.Compared to traditional fine-tuning methods, SWEET-RL delivers higher performance with less overfitting and greater generalization.Check outthe Paper, GitHub Page and Dataset.All credit for this research goes to the researchers of this project. Also,feel free to follow us onTwitterand dont forget to join our85k+ ML SubReddit. NikhilNikhil is an intern consultant at Marktechpost. He is pursuing an integrated dual degree in Materials at the Indian Institute of Technology, Kharagpur. Nikhil is an AI/ML enthusiast who is always researching applications in fields like biomaterials and biomedical science. With a strong background in Material Science, he is exploring new advancements and creating opportunities to contribute.Nikhilhttps://www.marktechpost.com/author/nikhil0980/OpenAI Introduced Advanced Audio Models gpt-4o-mini-tts, gpt-4o-transcribe, and gpt-4o-mini-transcribe: Enhancing Real-Time Speech Synthesis and Transcription Capabilities for DevelopersNikhilhttps://www.marktechpost.com/author/nikhil0980/How to Use SQL Databases with Python: A Beginner-Friendly TutorialNikhilhttps://www.marktechpost.com/author/nikhil0980/Cloning, Forking, and Merging Repositories on GitHub: A Beginners GuideNikhilhttps://www.marktechpost.com/author/nikhil0980/This AI Paper Introduces a Latent Token Approach: Enhancing LLM Reasoning Efficiency with VQ-VAE Compression

Sedona, Arizona, doesn't have many internet service providers. However, T-Mobile's fixed wireless service will be a reliable choice.

Threads' new update gives you more control, and will hopefully limit trolls and bots from spamming your posts.

It feels like open world games have been having a bit of an extended, existential crisis these past few years. Like there's a sense that something within the classic formula of big maps, question marks and to-be-coloured-in icons that's served us so well, back through The Witcher 3 and Skyrim to at least The Elder Scrolls 4: Oblivion, isn't quite working. That a change or evolution - if not outright ripping-up revolution - of some kind is necessary for the genre to thrive. But also, a bit of a problem: that the games wrestling with these bold new frontiers of mapmaking don't really know what that new, evolved form ought to be.I'm choosing to place the blame for this squarely with The Legend of Zelda: Breath of the Wild, a game so momentously influential in its design that it's really a clich now to even mention it. But mention it I must! Because as we all remember, Breath of the Wild was the first big-budget, third-person, open world action game in the post-Skyrim landscape to come out and do something properly different with its structure - and succeed in doing so. It did this by shedding systems of incrementally ever-increasing gear numbers for tiers of self-destructing twigs, and dropping the mass of map icons so heavily relied on by most of its contemporaries for something less prescriptive, more topographical. A map that drew your eye with its contoured details and curious formations, rather than literal waypoints, for an approach that we all decided in joyous union was much more artful.And from there it all got a bit messy. Open-worlders took different stabs at borrowing from Zelda, more often than not landing somewhat strangely on including a paraglider, of all things. No game summarised this more aptly than Assassin's Creed: Valhalla, the last full-sized Assassin's Creed offering from Ubisoft and in some ways perhaps also the strangest. Valhalla's approach to maps and discovery baffled and fascinated: beginning as a fogged-out wash of south-easterly English parchment, gradually this map unveiled its alternative to the Big Question Mark problem. We can't just stick a load of mystery boxes on there anymore, I'm assuming the discussion went, because it's all a bit too low-brow - so what do we do? Well, we keep them on there but replace them with glowing dots instead: silver for, erm, something, and gold for when there's rare loot. In searching for a reminder of Valhalla's map I found this Map Genie one from our sister site IGN, which feels apt. | Image credit: Eurogamer / IGN MapgenieThe result was a little murky, an actually quite fascinating idea, blending two almost diametrically opposed approaches to discovery into one - the icon-and-question-mark maximalism of prior Assassin's Creeds such as Odyssey with the wistful minimalism of Zelda - into something that ultimately landed on neither the former's convenience nor the latter's romance. So many games in this loosely post-Skyrim period have got themselves into such a tizz over how they implement their open worlds, as they've found themselves in everything from Halo and Gears of War to God of War and Call of Duty. So many more have become even more puzzled by Breath of the Wild, and this endless struggle between a will to make things painless for a player and, I've always suspected, a nagging belief that actually, getting your players to think about things actively might be best. Ubisoft, so brutally maligned for its commitment to a very specific, aptly ubiquitous open-world structure of tower-climbing and icon-revealing, has arguably suffered the most, caught in this tangle of opposing ideas and lost identities as a result.The issue was, and is, that the real problem is really only question mark-adjacent. The critique of this approach was really of the entire philosophy of Assassin's Creed games, and beyond that, the many open worlds, Ubisoft and otherwise, that both preceded and followed it. Namely: passivity. You play these games on autopilot, as exemplified in recent Assassin's Creeds so perfectly by the fact you could even set your horse to auto-run to your selected destination, a comfortable hands-free experience that let you blast yourself with a couple more dopamine hits from your phone in between crunching skulls, popping loot and cracking open chests.Your mindset as a player here, in games that woo you into passive mode, isn't to discover, investigate, experiment and enquire, as much as it is to be regularly presented with gameplay by the game in rhythmic, perfectly-timed intervals. For most, if not always all hints of friction to be lubricated into oblivion by all the soothing, massaging tools in the box. That means question marks for the nose-leading lures, yes, but also GTA-style sat-nav directions to get you there; pinned bullet-point objective reminders in the top-left; instant fast-travel; diegetic icons above all interactable characters' heads; chests that make sounds; loot that teleports itself to your magic stores when you forget to collect it; and, naturally, good old yellow paint on the climbable ledges. Each a kind of silent homage to the big, invisible Bioshock arrow in the sky. In other words: feed me; do not make me hunt - unless the hunting feels like feeding as well. Image credit: Eurogamer / UbisoftAgain, all of this is pretty well-worn discussion now. We had it with the great high-brow critics' sigh of relief when Breath of the Wild came out and all the (somewhat misplaced) hope it inspired for more inventive triple-A action-adventure designs (guilty). And in the simmering discontent at said yellow paint splashes whenever they come splattering (also guilty). But also, even in comments from quite brilliant game directors themselves, too."Travel is boring? That's not true. It's only an issue because your game is boring. All you have to do is make travel fun," Dragon's Dogma 2 director Hideaki Itsuno told IGN, when asked about fast travel early last year, damning an entire modern discipline in a sentence. "That's why you place things in the right location for players to discover, or come up with enemy appearance methods that create different experiences each time, or force players into blind situations where they don't know whether it's safe or not 10 metres in front of them." In time, the discussion has become distilled, as it always does, into a kind of analytical truism: hands-off, Breath of the Wild-style discovery good; Ubisoft-style hand-holding bad.Well, I disagree. Or really I somewhat disagree, but that doesn't sound as interesting. The hands-off approaches of modern Zelda games, of Dragon's Dogma 2, and the smaller-scale efforts influenced in that direction - think: Sable - are wonderful, yes. Bordering on genius in fact. I think they are in most senses 'better', requiring more sophistication and nuance in design, or at least more wilful engagement from you as a player - this isn't me going all relativism on you about how everything's just like, your opinion, man. But, also: it would be wrong to say there isn't some genius in there, somewhere, amongst the Ubisofts of this world.It's a tougher sell to describe, mind. Picture this: you're galloping through the tree-dappled hills of outer Osaka towards a distant landmark - the masterly, swooping tenshu of Osaka Castle, perhaps - when you spot a troubled villager by an overturned cart next to the road. You hop off for a quick chat and, a bit of conversational guesswork later, you blag some intel out of them on rumours of a stray dog nearby in need of comforting. The dog's in the other direction to the castle, you note on looking at the freshly-added marker on your increasingly thronging map, but it's a bit closer; the castle can wait. Image credit: Eurogamer / UbisoftSo off you go to find this dog, and a little quest ensues as you'd expect, and then you're up a hill and close to an eagle tower (we still got 'em!) that could do with unlocking, so you scurry off to do that, and now more of your foggy map's been cleared and more question marks appear, the last couple in the area, which bugs you. And so you think: better go clear off those question marks since I'm nearby! And those question marks lead to more quests, which are near more eagle towersThis stuff is the opposite of intentionality, the opposite of control, arguably the opposite even of empowerment, or fulfilment. When games like modern Assassin's Creeds and other Ubi fare get described as "junk food" there is more accuracy there than people realise - though arguably the better term for it is snack food. A good snack isn't fulfilling, after all. Good snacks are moreish! And they're moreish by design, engineered with ultra-processed precision to keep those orange-dusted fingers scrabbling back into the bowl.I referenced a brilliant blog last week on the ways games have been designed towards addiction, not just in the obvious gambling-adjacent methods but in the many video game-specific ones, from the brain tickle of levelling-up to the creaking open of a chest, and the way these sister industries aim not to keep you spending, but playing, through the constant trickle of mini-rewards. You can chuck map icons and question marks in there too - the same fizzing urge to tear open the sidequest-shaped booster pack and see what task lies within - but this only leads to another question. Is all this, actually, a bad thing?This is where it gets more interesting, and more complicated. The argument against is similar, but not identical, to the argument against those finger-staining snacks. It's an argument based on immediate harms, one, but also of a kind of systemic control wielded over you. Going back to that analogy: by engineering food in such a way as to be perpetually moreish, purposefully unfulfilling, you are being manipulated into eating and buying more and more of it. Bad for your health, yes, but also bad in a deeper kind of way, bad because of the numbing effect of it all, the smoothing away of any real, conscious choice, the images of zombified grazing it conjures; bad for your spirit. Image credit: Eurogamer / UbisoftSo it goes, this argument would say, for games like Assassin's Creed and the rest. To be sucked into this zen loop of discovering, following, discovering, following again is to be stupified by it. And so, one, there's that immediate harm - sitting for hours longer than intended is as lethal as anything in our diets, we're told, but hey we're all used to that now - and perhaps more pressingly the spiritual one. Here, in this temple of escapism, your senses are dulled, your awareness of and righteous outrage at the wider world and its endless injustices muted, your will to effect really meaningful change sapped away by the endless discovery loop. All the spare minutes you'd typically use with purpose instead tumbling into another hour spent climbing big, beautifully rendered digital trees and jumping off them again.Is this actually bad, then? Well, it depends. We've been dulling our senses for centuries, after all. If I weren't busy shanking nasty Samurai in Assassin's Creed: Shadows this past week, I might instead have been soaking in a hot bath, taking a little walk, or cracking open my Fellowship of the Ring 4K Blu-ray steelcase for the 247th time. Or maybe eating some crisps. We have a duty not to escape ourselves out of existence, absolutely, and the need to push back against it is absolutely heightened by the sheer volume of acutely tuned devices, systems and programs sucking all the attention out of our free time.But we also have a duty to rest between struggles, to treat ourselves, and to appreciate the artistry of a big, fat, dripping burger of a video game as well. This is all part of the cosmic mix. It's good we're asking questions about it, good we're drawing lines in the sand, good we're waking up to the ways we're being conditioned or unplugged. And also good that developers such as Ubisoft are becoming more comfortable with their place amongst all this, as purveyors of slightly baser kinds of fun. Image credit: Eurogamer / UbisoftWhere's the genius in all this then? Well, there's equal pleasure to be had in placing yourself at the top of a long and intricately crafted water slide, closing your eyes, letting its over-chlorinated waters wash over you, being whisked away down the spiralling tunnel of fun and giving yourself over to a cheap, simple, but thoroughly earned smile. As long as it's your choice going into it. There's a place for this kind of video game, and you don't need permission to let yourself love it. We just need balance. Osaka Castle can wait, after all, until I choose to tackle it on my own terms.

Arriving next weekBe sure to cast your votes in the poll below; but first, let's check out the box art designs themselves.North AmericaImage: Koei / LaunchboxBoth designs here share similarities; primarily the main key art showcasing Nobunaga in his iconic Samurai armour. Here, however, the colours are a bit darker, and everything just looks a tough classier than its Japanese counterpart. That said, the title itself is just a bit... meh. Not awful, but it kind of looks out of place, y'know?JapanImage: Koei / LaunchboxJapan's version is, as expected, displayed in a portrait orientation, so we get to see more of that awesome Nobunaga key art. In the corners, we've also got pixel art images of characters from the game, which is a nice contrast. Finally, the gold mixed with the red Japanese text work really nicely together.Which region got the best Nobunaga's Ambition box art? (76 votes)North America22%Japan78%Thanks for voting! We'll see you next time for another round of Box Art Brawl.Related GamesSee Also

Published March 23, 2025 6:00am EDT close Chinese robot's kung fu moves will make your jaw drop A humanoid robot has transformed from animble dancer to a martial arts master. In a stunning display of technological advancement, China's Unitree Robotics has unveiled its latest feat, a humanoid robot that can perform kung fu moves with astonishing precision and balance.The G1, Unitree's compact humanoid robot has transformed from animble dancer to a martial arts master, showcasing the rapid progress in robotics and artificial intelligence. G1 humanoid robot (Unitree Robotics)From virtual training to real-world masteryUnitree's approach to developing the G1's skills is as fascinating as the robot itself.STAY PROTECTED & INFORMED! GET SECURITY ALERTS & EXPERT TECH TIPS SIGN UP FOR KURTS THE CYBERGUY REPORT NOWThe process begins in a virtual environment using Nvidia's Isaac Simulator, whereby the robot learns complex behaviors before it even exists in physical form. This innovative method involves creating a digital twin of the humanoid robot that observes and learns from human actions using motion capture and video data.The behaviors are then refined through reinforcement learning in the virtual world. Subsequently, these acquired skills are transferred to the physical robot using a technique called Sim2Real, which seamlessly bridges the gap between simulated actions and real-world applications. G1 humanoid robot (Unitree Robotics)Kung fu mastery on displayIn its latest video demonstration, the G1 humanoid robot performs an array of impressive kung fu movements with remarkable balance and agility. The robot executes punches, roundhouse kicks and other complex martial arts techniques, showcasing its enhanced coordination and flexibility. With 23 degrees of freedom, the G1 demonstrates a level of dexterity that would make even Bruce Lee raise an eyebrow.WHAT IS ARTIFICIAL INTELLIGENCE (AI)? G1 humanoid robot (Unitree Robotics)Beyond martial arts: A versatile helperWhile the kung fu demonstration is undoubtedly eye-catching, Unitree envisions a broader role for its humanoid robots. The company positions the G1 as a versatile machine capable of handling challenging, repetitive tasks across various settings, including homes, factories and hospitals. This aligns with Unitree's vision of humanoid robots serving as useful companions in both work and daily life. G1 humanoid robot (Unitree Robotics)Open-source innovationTo further advance the natural movement of its humanoid robots, Unitree has released an open-source full-body dataset. This dataset, compatible with the G1, H1 and H1-2 models, enables the robots to perform human-like motions with improved flexibility and coordination. The dataset incorporates a redirection algorithm that optimizes the robots' movements, taking into account factors such as end posture constraints, joint positions and velocity limitations.GET FOX BUSINESS ON THE GO BY CLICKING HERE G1 humanoid robot (Unitree Robotics)The future of humanoid roboticsAs we witness the G1's transformation from adancing robot to a kung fu master, it's clear that the field of humanoid robotics is advancing at an unprecedented pace. The combination of sophisticated hardware, advanced AI algorithms and innovative training techniques like Sim2Real is pushing the boundaries of what these machines can achieve.While the demonstration of martial arts skills is impressive, it also raises questions about the future applications and implications of such advanced robotics. As these machines become increasingly capable of mimicking human movements and behaviors, we must consider both the potential benefits and the ethical considerations that come with this technology.Whether these machines will ultimately become our helpful companions or raise concerns about the future of human-robot interactions remains to be seen. One thing is certain: the field of humanoid robotics is evolving rapidly, and we're only beginning to scratch the surface of what's possible. G1 humanoid robot (Unitree Robotics)Kurt's key takeawaysIt's hard not to be amazed by Unitree's G1 humanoid robot, which has transformed from a nimble dancer to a kung fu master in a remarkably short time. This isn't just about cool martial arts moves. It's a glimpse into a future where robots could become our everyday helpers. But as we celebrate these advancements, we also need to think about what they mean for our relationship with technology.CLICK HERE TO GET THE FOX NEWS APPAs robots like Unitree's G1 become increasingly capable of mimicking human movements and behaviors, do you think we should be excited about the potential benefits or concerned about the potential risks of creating machines that can perform complex tasks, including martial arts? Let us know by writing us atCyberguy.com/ContactFor more of my tech tips and security alerts, subscribe to my free CyberGuy Report Newsletter by heading toCyberguy.com/NewsletterAsk Kurt a question or let us know what stories you'd like us to coverAlert:Malware steals bank cards and passwords from millions of devicesFollow Kurt on his social channelsAnswers to the most asked CyberGuy questions:New from Kurt:Copyright 2025 CyberGuy.com. All rights reserved. Kurt "CyberGuy" Knutsson is an award-winning tech journalist who has a deep love of technology, gear and gadgets that make life better with his contributions for Fox News & FOX Business beginning mornings on "FOX & Friends." Got a tech question? Get Kurts free CyberGuy Newsletter, share your voice, a story idea or comment at CyberGuy.com.