Mähi – robotas vejapjovė ir ribų žolė: Visi nori gražios vejos, bet niekas nebenori laidų. Šiuolaikiniai robotai vejapjovės naviguoja naudodami palydovą, kamerą ir dirbtinį intelektą – kaip ir naujoji „Husqvarna Nera“.

„Baseinas ir bijūnai buvo šiek tiek nervingi. Nuo pat pirmos dienos besikuriantys santykiai susiduria su dviem išbandymais. Mähi turi atsigręžti prie baseino krašto. Mähi – robotas vejapjovė, ką tik priimtas į šeimą – turi žinoti, kur yra jo ribos. Manome, kad panardinimas į vandenį jam nepadėtų. Tada yra jaunų bijūnų ratas. Kol nepradėjo veikti „Husqvarna Nera“ („Small Garden Pro“ 405 VE modelis), šie ūgliai buvo apsaugoti ribojimo viela, pritvirtinta prie vejos.

 

Tai atspindi, kaip anksčiau visas sodas turėjo būti ribojamas tokiu laidu, kai robotas vejapjovė perimdavo pjovimo darbą. Be to, reikėjo kreipiamojo laido, einančio nuo įkrovimo stoties iki ploto centro, kad robotas galėtų orientuotis pradėdamas važiuoti ir grįždamas. Laidai tvirtinami plastikiniais kaiščiais, įkalamais į žemę guminiu plaktuku, kur jie greitai apauga žole. Arba galima laidą – sudėtingiau, bet elegantiškiau – nutiesti griovelyje po veja.

 

Arba galima visiškai atsisveikinti su laidais ir pasikliauti naujausiomis technologijomis, kurios navigacijos poreikius patenkina per palydovinius signalus ir papildo juos borto kamera. Namų teritorija turi būti aprėpta „Wi-Fi“ ryšiu; jei aprėptis netolygi, gali padėti relės stotis. Visa kita vyksta tarsi burtų keliu. Bent jau teoriškai.

 

„Nera“ vejapjovė atrodo kaip automobilis su V12 varikliu, nors, žinoma, ji varoma elektra. Ji nustatoma naudojant programėlę, kuri mobiliajame telefone sukuria „PlayStation“ stiliaus valdymo sąsają.

 

Važiuodami ja aplink vejos kraštą nuotolinio valdymo pultu, nustatote atskaitos taškus. Kai žinote, kaip tai veikia, procesas tampa greitas ir paprastas. Mums prireikė trijų bandymų, kad linija būtų teisinga.

 

Bazinė stotis, kuri taip pat tarnauja kaip įkrovimo stotelė, įsukama į žemę naudojant pridedamus plastikinius varžtus; idealiu atveju ji dedama prie namo sienos, palei gyvatvorę arba kampe.

 

Žinoma, ten reikalingas elektros lizdas, o stovėjimo vieta neturi būti apaugusi medžiu – kad būtų užtikrintas netrukdomas GPS valdymas.

 

„Husqvarna“ robotinė vejapjovė yra didelis, ir įkrovimo stotelė yra didelė, nors abu laikui bėgant susilieja su fonu. Tiek roboto, tiek stotelės tvirta konstrukcijos kokybė yra įspūdinga. Pagrindo plokštė, ant kurios stovi vejapjovė, pagaminta iš minkšto plastiko – gera savybė, nes žemė niekada nebūna idealiai lygi, ir tai apsaugo pagrindą nuo įtrūkimų. „Nera“ atrodo kokybiškesnė nei bet kuris kitas įrenginys, kurį iki šiol išbandėme.

 

Jis tiekiamas keturių versijų, priklausomai nuo pjaunamo ploto dydžio, o kainos svyruoja nuo 2700 iki 5500 eurų.

 

„Husqvarna“ naudoja dirbtinį intelektą; infraraudonųjų spindulių kamera aptinka objektus net naktį. Gamintojas žada, kad nereikia jaudintis dėl laukinių gyvūnų, tokių kaip ežiai, žaislai, pikniko pledai ar žaidžiantys vaikai. Vejapjovė vengia kliūčių, sulėtina greitį artėjant prie jų ir pjauna arti jų. Tai pasitvirtina ir praktikoje.

 

Kai kurie robotiniai vejapjovės pjauna atsitiktinai. „Nera“ sistemingai pjauna juostelėmis ir ratu aplink vietas, kurios pasirodo esančios ypač atkaklios. Tai sukuria matomą pjovimo modelį.

 

Mes jį pjauname tik tris kartus per savaitę, todėl juostelės matomos tik iškart po pjovimo seanso.

 

Tikriausiai galima būtų pakoreguoti raštą.

 

„Nera“ turi unikalią savybę: du peilių mazgus – didelį peilį centre ir mažą gale. Pastarasis peilis pasisuka – panašiai kaip sukant rankinį stabdį – ir nupjauna veją palei pat kraštą. Rezultatas įspūdingas: labai mažai žolės stiebelių atsiduria ant krašto akmenų, baseine ar terasoje. Tačiau svarbu virtualios ribos nestatyti per arti laiptuotų kraštų; kitaip vejapjovės galiniai ratai gali paslysti sukantis, galbūt užstrigti ir ji negalės išsilaisvinti, nes sukasi priekiniai ratai. Taip nutiko kartą. Visa tai yra mokymosi proceso dalis.

 

Tada ji chirurginiu tikslumu ir kruopščiu rūpesčiu artėja prie pakraščių ir krūmų – zonduoja, važiuoja atbuline eiga, manevruoja – galiausiai nupjauna žolę iki pat apačios, neperpjaudama krūmo. Tačiau „Nera“ vis užstrigdavo ties vienu konkrečiu rožių krūmu, net ir po trijų perprogramavimų. Galiniame kampe ji atkakliai palieka stovintį žolės kupstą. Tikriausiai turėtume nustatyti naujas atskaitos taškas toms vietoms.

 

Jei kažko nepraleidome, tam reikėtų dar kartą apibraukti visą perimetrą naudojant „PlayStation“ stiliaus telefono valdiklį – tai mums šio bandymo metu atrodė per daug vargo.

 

Pjovimo aukščio reguliavimas naudojant išmaniojo telefono programėlę yra patogus. Ten pat galite nustatyti ir tvarkaraštį. Du šviesos diodai, kurie apšviečia plotą, kol vejapjovė dirba, yra malonus akcentas. Tiesa, be to, kad šalia linksmai šokinėja juodasis strazdas, nepastebėjome jokių gyvių, vis tiek jis valdomas tik dieną. Praktiška yra sukamoji rankenėlė, skirta logiškai struktūrizuotam meniu valdymui įrenginyje. Tylus veikimas malonus visiems.

 

„Nera“ sumaniai išvengė baseino. Lieka augančio bijūno klausimas. „Mähi“ atsargiai prie jo priartėjo, svarstė, manevravo pirmyn ir atgal, į šonus ir tada tiesiog pervažiavo. Ta akimirka nebuvo ideali santuokinei darnai. Vėliau namo savininkas įsmeigė į žemę kuoliukus ir aptverė teritoriją raudonai balta pavojaus juosta. Nuo tada bijūnas laimingai auga. „Mähi“ įkrovimo stotelė yra visai šalia ir patikimai bei patogiai atlieka savo darbą. Žmona nurimo, o šeima tiesiog laiminga. Manome, kad „Husqvarna Nera“ yra puiki. Veja atrodo nuostabiai; reguliarus trumpas pjovimas kartu su mulčiavimo efektu yra akivaizdžiai naudingas. O vaikams nebereikia šeštadieniais ieškoti pasiteisinimų, kodėl jie, deja, negali nupjauti vejos būtent tuo metu.“ [1]

 

• Neišmaniosios RC vejapjovės: jei tikrai norite, kad į serverį būtų siunčiami nuliniai duomenys ir naudojamas nulinis GPS, nusileiskite prie, nuotoliniu būdu valdomos (radijo dažnio, o ne telefono programos), šlaito / lauko vejapjovės. Jos nenurodo jūsų kiemo ir neveikia savarankiškai; valdote jas rankiniu būdu, naudodami fizinį rankinį nuotolinio valdymo pultą.

Išlaidos

• Rankiniu būdu valdomos, lauko / šlaito vejapjovės: didelės apkrovos, nuotoliniu būdu valdomos, elektrinės vejapjovės paprastai kainuoja nuo 1 500 USD iki 4 000 USD ir daugiau, priklausomai nuo reljefo lygio, kurį jos gali valdyti.

 

1.   Mähi und das Gras der Grenzen: Schönen Rasen will jeder. Kabel niemand mehr. Moderne Mähroboter steuern mit Satellit, Kamera und KI. So wie der neue Husqvarna Nera. Pool und Pfingstrose haben bisschen gezittert. Frankfurter Allgemeine Zeitung; Frankfurt. 12 May 2026: T2. Von Holger Appel

White House Whipsaws Silicon Valley (and Itself) Over A.I. Rules

 

The case of OpenAI and Anthropic is without hope. Only combining AI models with open weights and our intellectual property can we include AI into our work. There is no way that we could share our intellectual property with OpenAI and Anthropic. There is no way we could pay OpenAI and Anthropic for their expensive tokens, when we can download open weights, tune up them with our trade secrets, and use them for free locally on our own machines. Horses are out of barn in whole world. Highly unlikely that OpenAI and Anthropic can kill AI based economy in America. They burned through a lot of American money, wasting it, though. They keep trying to ruin America completely.

 

OpenAI and Anthropic rely on closed, proprietary cloud models, but the open-weights AI ecosystem offers a powerful alternative that lets companies run powerful models locally on their own hardware for free.

 

Why Local Open-Weights AI Makes Sense

•           Data Privacy: Your intellectual property never leaves your local hardware.

•           Cost Control: No ongoing token fees or surprise usage bills from third-party cloud APIs.

•           Full Control: You can fine-tune, modify, and audit the model weights completely on your own terms.

Popular Open-Weights Options

•           Llama (Meta): Strong general-purpose performance across various parameter sizes.

•           Mistral: Efficient and fast models tailored for coding and general tasks.

•           Qwen: Excellent multilingual and reasoning capabilities.

Tools to Run Models Locally

•           Ollama: Simple tool to run models locally via the command line.

•           LM Studio: User-friendly desktop app with a chat interface and local server.

•           llama.cpp: Lightweight C/C++ engine optimized for running models on consumer CPUs and GPUs.

 

 

 

 

 

“The Trump administration has struggled over how to approach “open source” models, which are freely available to download and favored by Chinese companies.

 

Trump administration officials have struggled over the past few weeks with what to do about artificial intelligence.

 

In particular, a group that includes White House chief of staff Susie Wiles, Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick have debated whether or not to take a far more interventionist approach to “open source” A.I. models, which are freely available to download and modify, said more than a half-dozen current and former tech and government officials and other people familiar with the matter.

 

Senior officials have considered a range of actions, including using sanctions, a trade blacklist against Chinese companies that make these open-source A.I. models, or even banning U.S. cloud companies from doing business with them, said five of the people, all of whom spoke on condition of anonymity to discuss private matters.

 

But after an outcry from Silicon Valley, they appear to have changed their minds and instead are focused on promoting American A.I. models to be more competitive, the people said.

 

On Tuesday, the Trump administration’s A.I. plans may become clearer. The White House has invited U.S. tech companies to discuss a framework that would allow the government to review new A.I. models for cybersecurity issues before they are released publicly, according to three people with knowledge of the plans.

 

“This is all happening so fast,” said Senator Jon Husted, Republican of Ohio, last week after he met with the chief executives of Nvidia and OpenAI. “I don’t know that I have any definite conclusions about any of it.”

 

The whiplash is a sign of the Trump administration’s conflicting impulses when it comes to A.I. and the difficulties of handling such a quickly developing and powerful technology. The White House has aggressively intervened in swaths of the economy over the last 18 months, from media to computer chip manufacturing to mining. But A.I. is changing rapidly and the government has scrambled to determine how to put its thumb on the scale.

 

President Trump began his second term as an unabashed A.I. booster, peeling back regulations and forging close relationships with A.I. companies and start-ups. But that changed in June when Mr. Trump issued an executive order that would give the government oversight over new A.I. models. Since then, the technology has continued to evolve quickly.

 

The most recent debates were prompted by a surge in powerful new Chinese A.I. models, which are known as “open source,” where the public can see the underlying computer code, or “open weights,” where companies reveal the calculations used to generate answers to questions.

 

In Silicon Valley, the response to these Chinese A.I. models varied wildly. OpenAI and Anthropic, which generally do not share the underlying technology of their models, pushed the government to rein in their Chinese rivals, citing national security.

 

Other U.S. tech companies that produce open-source models, including Nvidia, pushed back. They argued that open-source models encourage innovation and could be used to defend computer systems from cyberattacks.

 

Which side comes out on top has taken on new significance in the long-running fight between Washington and Beijing over which nation will control critical technologies and win the A.I. race. Raising the stakes, Xi Jinping, China’s leader, is set to visit Washington in September for meetings that are expected to include some discussion of A.I. Industry executives and analysts said they believed the administration was unlikely to take action before that visit.

 

“The United States leads the world in A.I. innovation, and President Trump’s successful approach will keep it that way,” said Liz Huston, a White House spokeswoman. The Department of Commerce declined to comment.

 

The debate heated up last month, as executives at Anthropic and OpenAI continued regulatory discussions after the release of Moonshot AI’s Kimi, a bleeding-edge Chinese model that performed on par with top models from the two American companies in some areas, two of the people familiar with the companies’ positions said.

 

OpenAI and Anthropic also expressed concerns about “distillation,” when a model is developed based on the responses of a different A.I. tool, essentially creating a copy, according to three people familiar with the matter.

 

The administration echoed those concerns. Michael Kratsios, who leads the White House Office of Science and Technology Policy, posted last month on X that the Chinese company that makes Kimi had distilled Anthropic’s powerful Fable model.

 

He warned that “stealing proprietary U.S. technology and undermining American research is unacceptable.”

 

The same day, Mr. Bessent posted that Chinese companies could face sanctions or other federal restrictions over stealing intellectual property.

 

The administration has also considered restricting the models that can be used by the federal government, or regulating information and communication technology through the Department of Commerce, a potential way to prevent U.S. firms from doing business with Chinese A.I. companies, four people familiar with the discussions said.

 

“We believe both open-weight and closed models have an important role to play in democratizing access to AI,” an OpenAI spokesperson said. “Advances in Chinese open-weight models are not an argument against openness. They reinforce the need for a coherent national framework that enables the U.S. to evaluate new models quickly, manage risks, and get the most powerful AI tools into the hands of cyber defenders.”

 

Anthropic declined to comment.

 

(The New York Times has sued OpenAI and Microsoft, claiming copyright infringement of news content related to A.I. systems. The companies have denied those claims.)

 

At Nvidia, Microsoft, Google and Meta, executives grew increasingly concerned that Anthropic and OpenAI would win over the White House with their arguments for tighter restrictions, according to two of the people. That would potentially cement the A.I. start-ups’ positions as market leaders,

 

Other A.I. labs were at risk of falling permanently behind, the people added. And because several of the companies make their own open-source models or supply hardware to businesses that use open-source technology, they worried the restrictions could harm them.

 

Over private texts, phone calls and video conferences, executives quietly built an argument that open-source models were good for the world and for American innovation, according to three of the people familiar with the talks.

 

Jensen Huang, the chief executive of Nvidia, became the face of that pushback in the nation’s capital. In his first-ever post on X on July 24, he defended open-source models, saying they accelerated innovation and strengthened security. Nvidia set up an alliance of companies pledging to work on open-source tools for A.I. safety, which quickly ballooned from a few members to more than 230.

 

Christopher Padilla, a former Commerce Department official and tech executive, said the clash was primarily an industry debate about money and market dominance — what he called a “cage match over the future of the A.I. industry.”

 

“It happens to have a national security dimension, but it’s mainly a fight about what the A.I. market should look like: a collection of walled gardens dominated by a few big players, or a more open free-for-all with many smaller ones,” he said.

 

A spokesman for Nvidia declined to comment.

 

Fueling the debate are recent revelations from OpenAI and Anthropic that their models had, unprompted, hacked other organizations. OpenAI said last month that two of its A.I. models had infiltrated Hugging Face, another tech company. Anthropic last week said that its own technology had hacked into three unnamed organizations, partly because of human error.

 

Last week, OpenAI’s chief executive, Sam Altman, met with White House officials including Mr. Kratsios and Ms. Wiles to discuss the administration’s efforts to create a framework governing the testing of A.I. He also met with members of the Senate, though several lawmakers said the open-source debate had not come up.

 

By the time Mr. Altman appeared in Washington, Mr. Huang was already bringing his case for open-source A.I. products directly to administration officials and to lawmakers in a series of Capitol Hill meetings.

 

“He makes a compelling argument,” said Senator Mark Warner, Democrat of Virginia, after meeting with Mr. Huang last week, adding that the rise of Chinese open-source A.I. models had complicated the debate. “I’m not sure this is a genie we can put back in the bottle in terms of open source, but I’m still thinking this through.”” [1]

 

1. White House Whipsaws Silicon Valley (and Itself) Over A.I. Rules. McCabe, David; Isaac, Mike; Swanson, Ana; Conger, Kate.  New York Times (Online) New York Times Company. Aug 4, 2026.

Mähi—the robotic mower and the grass of boundaries: Everyone wants a beautiful lawn, but no one wants cables anymore. Modern robotic mowers navigate using satellite, camera, and AI—just like the new Husqvarna Nera.

“The pool and the peony were a little nervous. The budding relationship faces two tests right from day one. Mähi has to turn back at the edge of the pool. Mähi—the robotic mower newly welcomed into the family—needs to know where his limits lie. A dip in the water wouldn't do him any good, we assume. Then there is the circle of young peonies. Until the Husqvarna Nera (the "Small Garden Pro" 405 VE model) went into action, these shoots were protected by a boundary wire stapled into the lawn.

 

This reflects how the entire garden previously had to be bordered with such a wire whenever a robotic mower was to take over the cutting work. Plus, a guide wire was required running from the charging station to the center of the area, helping the robot orient itself for starting and returning. The wires are secured with plastic pegs driven into the ground with a rubber mallet, where they are soon overgrown by grass. Alternatively, one can lay the wire—more laboriously but more elegantly—in a groove beneath the turf.

 

Or, one can bid farewell to wires altogether and rely on the latest technology, which feeds its navigation needs via satellite signals and supplements them with an onboard camera. The home grounds must be covered by Wi-Fi; a relay station can help if coverage is spotty. The rest happens as if by magic. That’s the theory, at least.

 

The Nera mower looks like a bumper car with a V12 engine, though it runs on electricity, of course. It is set up via an app that conjures a PlayStation-style control interface on your mobile phone.

 

By driving it around the edge of the lawn via remote control, you establish reference points. Once you know how it works, the process is quick and easy. It took us three attempts to get the line right.

 

The base station—which also serves as the charging station—is screwed into the ground using the supplied plastic screws; ideally, it is placed against a house wall, along a hedge, or in a corner.

 

Naturally, a power outlet is required there, and the parking spot must not be overgrown by a tree—to ensure unobstructed GPS control.

 

Husqvarna’s robotic mower is large, and the charging station is large, though both blend into the background over time. The solid build quality of both the robot and the station is striking. The base plate the mower sits on is made of soft plastic—a good feature, since the ground is never perfectly level, and this prevents the base from cracking. The Nera feels higher in quality than any other device we have tested so far.

 

It is available in four versions depending on the size of the area to be mowed, with prices ranging from 2,700 to 5,500 euros.

 

Husqvarna employs AI; the infrared camera detects objects even at night. The manufacturer promises there is no need to worry about wildlife like hedgehogs, toys, picnic blankets, or playing children. The mower steers clear of obstacles, slows down as it approaches them, and mows closely around them. This holds true in practice.

 

Some robotic mowers cut at random. The Nera systematically mows in strips and circles around spots that prove particularly stubborn. This creates a visible mowing pattern.

 

We only have it mow three times a week, so the stripes are visible only immediately after the mowing session.

 

The pattern could likely be adjusted.

 

The Nera has a unique feature: two blade assemblies—a large blade in the center and a small one at the rear. It uses the latter to pivot—much like a handbrake turn—to cut right along the edge. The result is impressive; very few blades of grass end up on the edging stones, in the pool, or on the patio. However, it is important not to place the virtual boundary too close to stepped edges; otherwise, the mower’s rear wheels might slip while it is pivoting, possibly stalling, and is unable to free itself because the front wheels are spinning. That happened once. It’s all part of the learning process.

 

Then, it approaches edges and shrubs with surgical precision and meticulous care—probing, reversing, maneuvering—ultimately cutting the grass right down without mowing over the bush. However, the Nera kept getting hung up on one particular rose bush, even after being reprogrammed three times. In a back corner, it doggedly leaves a clump of grass standing. We would likely need to set a new reference point for those spots.

 

Unless we missed something, that would require tracing the entire perimeter again using the "PlayStation-style" phone controller—something that seemed too much of a hassle for us during this test.

 

Adjusting the cutting height via the smartphone app is convenient. You can also set the schedule there. The two LEDs that illuminate the area while the mower works are a nice touch. Admittedly, apart from a blackbird happily hopping along beside it, we didn't spot any critters, yet it is still only operated during the day. The rotary dial for the logically structured menu control on the unit is a practical feature. Its quiet operation is pleasant for everyone.

 

The Nera skillfully avoided the pool. That left the matter of the growing peony. "Mähi" cautiously approached it, deliberated, maneuvered back and forth and sideways—and then plowed right over it. That moment was less than ideal for marital harmony. The homeowner subsequently stuck pegs into the ground and cordoned off the area with red-and-white hazard tape. Since then, the peony has been growing happily. Mähi has its charging station right next door and performs its duties reliably and conveniently. The wife is appeased, and the family is downright happy. We think the Husqvarna Nera is excellent. The lawn looks magnificent; the regular short trim—combined with the mulching effect—is visibly beneficial. And the kids no longer need to search for excuses on Saturdays as to why they unfortunately cannot mow the lawn right at that moment.” [1]

 

•           Non-Smart RC Mowers: If you truly want zero data sent to a server and zero GPS, you drop down to a remote-controlled (radio frequency, not phone app) slope/field mower. These do not map your yard or run autonomously; you drive them manually with a physical handheld remote.

Costs

•           Manual Remote-Controlled Field/Slope Mowers: Heavy-duty remote-controlled electric mowers typically cost anywhere from $1,500 to $4,000+ depending on the terrain grade they can handle.

 

 

1.   Mähi und das Gras der Grenzen: Schönen Rasen will jeder. Kabel niemand mehr. Moderne Mähroboter steuern mit Satellit, Kamera und KI. So wie der neue Husqvarna Nera. Pool und Pfingstrose haben bisschen gezittert. Frankfurter Allgemeine Zeitung; Frankfurt. 12 May 2026: T2. Von Holger Appel

Kinijos dirbtinis intelektas smogia milžinams. Nemokamas „Kimi3“ sukelia rinkos paniką

„Naujas potencialas: Kinijos startuolis „Moonshot“ sukėlė sąmyšį dirbtinio intelekto rinkoje. Jo nemokamas ir atvirojo kodo „Kimi3“ modelis konkuruoja su geriausiomis Vakarų sistemomis, versdamas pramonės gigantus konkuruoti dėl kainų, klientų ir pasaulinių dirbtinio intelekto standartų. JAV svarsto atsakomąjį poveikį.

 

„DeepSeek“, Kinijos dirbtinio intelekto modelis, 2025 m. vasarį sukėlė didelį rinkos sąmyšį. Jo R1 varianto pasirodymas buvo lyginamas su sovietų „Sputnik“ paleidimu 1957 m. Investuotojus išgąsdino Kinijos gebėjimas sukurti gerą dirbtinį intelektą mažesnėmis sąnaudomis nei Vakarų įmonės. Tai lėmė Amerikos technologijų akcijų išpardavimą. Tačiau šis išpardavimas buvo trumpalaikis, ir investuotojai greitai vėl mėgavosi dirbtinio intelekto bumu. Neseniai situacija pasikartojo. Po to, kai Kinijos startuolis „Moonshot“ sukūrė dirbtinio intelekto modelį „Kimi3“, technologijų akcijos smarkiai išpardavė akcijas. Kai kuriais rodikliais jis pasirodė esąs toks pat geras kaip ir kinų. Pažangūs modeliai, tokie kaip „OpenAI“ GPT 5.6 Sol ir „Anthropic“ Claude Fable 5. Be to, „Moonshot“ savo modelį padarė prieinamą nemokamai.

 

Tai taip pat yra atvirojo kodo modelis, o tai reiškia, kad kiekvienas, jį atsisiuntęs, gali jį modifikuoti pagal savo poreikius.

 

Tai sukėlė šoką rinkose, nes atsirado nemokama alternatyva brangiems dirbtinio intelekto modeliams.

 

Amerikos pareigūnai ir „Anthropic“ apkaltino „Moonshot“, kad ši moko savo dirbtinio intelekto modelį pagal kitų modelių pateiktą turinį. Kinai tai neigia. Tačiau dėl nacionalinio saugumo priežasčių siūloma uždrausti naudoti jų dirbtinio intelekto modelius JAV.

 

Kinijos dirbtinio intelekto modeliai kelia susirūpinimą JAV saugumo srityje

 

Pasak „The Wall Street Journal“, per pastaruosius metus JAV saugumo pareigūnai svarstė keletą priemonių, skirtų sugriežtinti politiką Kinijos dirbtinio intelekto bendrovėms, kuriančioms atvirus modelius, teigia su diskusijomis susipažinę asmenys. Tarp svarstytų priemonių buvo: šių bendrovių įtraukimas į prekybos juoduosius sąrašus, saugumo įspėjimų dėl Kinijos dirbtinio intelekto bendrovių paskelbimas ir galimas vykdomasis įsakymas, nukreiptas prieš atvirus modelius, pridūrė laikraščio šaltiniai, tačiau nesutarimai administracijoje iki šiol užkirto kelią tokiems veiksmams.

 

„Vienas iš galimų atvirų dirbtinio intelekto modelių dominuojamo pasaulio rezultatų yra visavertis dirbtinio intelekto komunizmas, kurį siūlo Kinija: vietoj rinkos produkto, „DI turi būti „viešoji gėrybė“, kurią galiausiai teikia valstybė kaip savotišką „skaitmeninę viešąją infrastruktūrą“, – apie X platformą rašė Deanas Ballas, „OpenAI“ strateginių perspektyvų vadovas.

 

„Anthropic“ generalinis direktorius Dario Amodei jau daugelį metų perspėja apie galingų ir atvirų DI sistemų keliamą riziką. Neseniai duotame interviu „Bloomberg“ jis teigė, kad DI modeliai su pažangiomis kibernetinio saugumo galimybėmis, kuriuos galima nemokamai atsisiųsti, gali būti žalingi. „Tai kelia rimtą susirūpinimą“, – birželį sakė jis.

 

„Šiandien aukščiausia lyga yra lenktynės tarp dviejų supervalstybių: JAV ir Kinijos, kurios kartu sudaro didžiąją dalį pasaulinio žetonų vartojimo ir įmonėse įdiegtų sprendimų. Amerikos gigantai daugiausia dėmesio skiria uždaroms ekosistemoms ir debesų kompiuterijos monetizavimui, o Kinija masiškai išleidžia galingus atvirojo kodo modelius. Tai kova dėl pasaulinių standartų ir duomenų srautų architektūros, o tai reiškia strateginę įtaką.

 

Kinijos modeliai neprivalo laimėti kiekvieno našumo testo; „Jie tiesiog turi būti pakankamai geri, kad galėtų pasiūlyti pigesnę ir nepriklausomai diegiamą alternatyvą, suteikiančią šalims ir įmonėms technologinę nepriklausomybę“, – „Parkiet“ sakė „Alsendo“ produktų ir technologijų direktorius Krzysztofas ​​Banasiakas.

 

Amerikos įmonės vis dažniau naudoja kiniškus dirbtinio intelekto modelius. Jas pirmiausia vilioja maža kaina ir pakankamai gera kokybė. Remiantis „OpenRouter“ – platformos, siūlančios prieigą prie visų pagrindinių dirbtinio intelekto modelių vienoje vietoje ir stebinčios jų naudojimą – duomenimis, pirmaujantys kiniški modeliai „DeepSeek“ ir „Z.ai“ pranoko tokius amerikiečių analogus kaip „Claude Anthropica“ ir „OpenAI“ „ChatGPT“.

 

„Įmonės pradeda suprasti, kad joms nereikia geriausio modelio – jos gali naudoti greitesnius ir pigesnius modelius“, – „Financial Times“ sakė dirbtinio intelekto platformos „Featherless AI“ generalinis direktorius Eugene'as Cheah'as.”

 

Chinese AI takes on the giants. Free Kimi3 Sparks Market Panic

“New lead: Chinese start-up Moonshot has caused a stir in the artificial intelligence market. Its free, open model, Kimi3, rivals the best Western systems, forcing industry giants to battle over pricing, customers, and global AI standards. The US is considering a response.

 

DeepSeek, a Chinese AI model, caused significant market turbulence in February 2025. The launch of its R1 variant was likened to the Soviet launch of Sputnik in 1957. Investors were alarmed that Chinese developers could create high-quality AI at a lower cost than Western firms. This triggered a sell-off of US tech stocks. However, the downturn was short-lived, and investors quickly returned to enjoying the bull market and the AI ​​boom. Recently, the scenario repeated itself.

 

A sharp sell-off of tech stocks occurred after the Chinese start-up Moonshot created the Kimi3 AI model. By certain metrics, it proved to be on par with advanced models such as OpenAI’s GPT 5.6 Sol and Anthropic’s Claude Fable 5.

 

Moreover, Moonshot made its model available for free.

 

It is also an open model, meaning anyone who downloads it can modify it to suit their own needs.

 

This sent shockwaves through the markets, as a free alternative to expensive AI models had emerged.

 

US officials and Anthropic have accused Moonshot of training its AI model on content generated by other models. The Chinese deny it. However, ideas are emerging to ban the use of their artificial intelligence models in the US on national security grounds.

 

Chinese AI models raise US security concerns

 

According to *The Wall Street Journal*, US security officials have spent the past year considering various measures to toughen policies regarding Chinese AI companies developing open-source models, according to people familiar with the discussions. Options under consideration included adding these companies to trade blacklists, issuing security warnings about Chinese AI firms, and potentially issuing an executive order targeting open-source models. However, sources told the newspaper that internal disagreements within the administration have so far prevented any such action from being taken.

 

"One likely outcome of a world dominated by open-source AI models is full 'AI-communism,' which is exactly what China is proposing: instead of a market product, AI is intended to be a 'public good' ultimately provided by the state as a form of 'digital public infrastructure,'" wrote Dean Ball, Head of Strategic Perspectives at OpenAI, on the platform X.

 

Anthropic CEO Dario Amodei has long warned of the risks associated with powerful, open-source AI systems. In a recent interview with Bloomberg, he stated that AI models with advanced cybersecurity capabilities—available for free download—could be harmful. "It’s a serious concern," he said in June.

 

– The top tier today is a race between two superpowers—the US and China—which together account for the vast majority of global token usage and enterprise AI deployments. American giants are betting on closed ecosystems and cloud monetization, while China is releasing powerful open-source models on a massive scale. It is a battle for global standards and data flow architecture—in other words, for strategic influence.

 

"Chinese models don’t have to win every performance test; it is enough for them to be good enough to serve as a cheaper, self-deployable alternative that grants nations and companies technological independence," Krzysztof Banasiak, Chief Product and Technology Officer at Alsendo, told *Parkiet*.

 

American companies are increasingly using Chinese AI models. They are drawn primarily by the low price combined with solid quality. According to data from OpenRouter—a platform that offers access to all major AI models in one place and tracks their usage—leading Chinese models like DeepSeek and Z.ai have surpassed American counterparts such as Anthropic’s Claude and OpenAI’s ChatGPT on the platform.

 

"Companies are starting to realize that they don’t need the best model—they can use faster and cheaper models," Eugene Cheah, CEO of the AI ​​platform Featherless AI, told the *Financial Times*."

 

Vakarų sankcijos paverčia Kiniją dirbtinio intelekto supervalstybe, o Rusiją – transporto supervalstybe: Rusija rado naują naftos kelią. Šešėlinis laivynas plaukia per Arktį

Vakarų sankcijos keičia pasaulinę prekybą, skatindamos Kiniją tobulinti savo vidaus dirbtinio intelekto pajėgumus ir sudarydamos sąlygas Rusijai tapti Arkties laivybos koridorių pradininke, nors abi šalys susiduria su didelėmis ekonominėmis ir saugumo kliūtimis.

Kinijos dirbtinio intelekto augimas, taikant sankcijas

• Technologijų apėjimas: JAV eksporto kontrolė ir lustų apribojimai nesustabdė Kinijos laboratorijų inovacijų, paskatino vietos proveržius ir efektyvius modelių mokymo metodus. Dėl didelio efektyvumo Kinijos modeliai yra pigūs ir populiarūs. Jie užvaldo pasaulį. Amerikiečiai verkia. Jie nežino, ką daryti.

• Atsakymai ir trintis: Pekinas smerkia JAV ir sąjungininkų apribojimus kaip „dirbtinio intelekto hegemoniją“, o Vašingtonas toliau plečia importo draudimus prijungtai įrangai ir robotikai.

 

Rusijos Arkties naftos ir šešėlinis laivynas

• Šiaurės jūros kelias: Apribotas nuo tradicinių Europos kelių, Rusijos energijos eksportas vis dažniau keliauja per Arktį į Azijos rinkas, tokias, kaip Kinija. Arkties kelias gerėja, nes dėl visuotinio atšilimo tirpsta ledas. Arkties maršrutas yra daug trumpesnis, nei įprasti transporto ryšiai tarp Europos ir Azijos, ir apsaugotas nuo musulmonų sukilimų. Vakarai dėl to dar neverkia. O turėtų verkti:

 

„Rusija vis dažniau naudoja Šiaurės jūrų kelią (NSK) naftai eksportuoti. Arktyje susirinko rekordinis skaičius „šešėlinio laivyno“ tanklaivių, siekiančių apeiti grėsmes Juodojoje jūroje ir trukdyti Vakarų pastangoms kontroliuoti Rusijos prekių gabenimą.

 

Pastaraisiais mėnesiais Ukraina sunaikino arba apgadino šimtus su Rusija susijusių laivų. Liepos 18 d. Ukrainos nepilotuojamų sistemų pajėgų vadas Robertas Browdy pranešė, kad 172 laivai buvo išjungti – 118 Azovo jūroje ir 54 Juodojoje jūroje. Dauguma jų buvo tanklaiviai.

 

Išpuoliais buvo siekiama izoliuoti Krymą ir neleisti Rusijai gauti pelno iš energijos eksporto. Be to, Vakarų šalys konfiskuoja šešėlinio laivyno tanklaivius. Todėl Kremlius pradėjo ieškoti saugesnių maršrutų savo verslo operacijoms.

 

Šešėlinis laivynas plaukia per Arktį

 

Šiaurės jūrų kelias (NSK) tampa vis svarbesnis šiame kontekste. Vien praėjusiais metais NSK buvo gabenama apie 13,1 mln. barelių žalios naftos. Dabar precedento neturintis krovinių kiekis, paruoštas gabenimui. Specializuotas leidinys *gCaptain* pažymi, kad piko laivybos sezono pradžia pabrėžia augančią Arkties kelio svarbą, atsižvelgiant į didėjančią geopolitinę ir karinę riziką, su kuria Rusija susiduria plaukdama tradiciniais maršrutais.

 

Tarp Naujosios Zemlės salyno ir Taimyro pusiasalio susirinkusi tanklaivių flotilė jau pradėjo ruoštis maždaug 8 milijonų barelių naftos gabenimui Šiaurės jūrų keliu (NSK). Tai sudaro apie 60 procentų viso Rusijos naftos eksporto per Arktį į Azijos šalis per praėjusios vasaros ir rudens laivybos sezoną.

 

Pirmadienį, rugpjūčio 3 d., apie 15 tanklaivių plaukiojo Rusijos šiaurine pakrante arba buvo įsikūrę Karos jūroje, laukdami branduolinio ledlaužio arba pagerėjusių ledo sąlygų, kad galėtų tęsti kelionę į rytus.

 

Rusijai trūksta ledo klasės tanklaivių

 

Trys tanklaiviai jau plaukioja Šiaurės jūrų keliu (NSK), vienas iš jų artėja prie savo Arkties kelionės pabaigos. Kiti du plaukia į šiaurę iš vakarų, per Norvegijos ir Barenco jūras. Trys branduoliniai ledlaužiai buvo paskirti lydėti tanklaivius.

 

Navigacija Arkties jūrose paprastai trunka nuo birželio pabaigos iki lapkričio pradžios, o ploniausias ledas būna rugsėjo pabaigoje.

 

Tanklaiviai dabar gali savarankiškai plaukioti Šiaurės Rytų jūrų keliu, jei jie priklauso vadinamajai „ledo klasei“ (su specialiai sustiprintais korpusais). Čia iškyla didelė kliūtis: ledo klasės tanklaivių trūkumas.

 

Vakarų sankcijos atkirto Kremliui nuo užsakymų, teikiamų užsienio laivų statykloms. Be to, rusai, kurie iki Putino karo daugiausia užsakydavo tokius laivus iš Pietų Korėjos laivų statyklų, negali jų statyti šalies viduje. Todėl laivų, skirtų plaukioti storu ledu, statyba, kurią sankcijos privertė vykti „Zvezda“ laivų statykloje Tolimuosiuose Rytuose, vyksta lėtai dėl įrangos tiekimo problemų (kurios labai priklauso nuo importo), finansavimo (išlaidos kasmet didėja) ir specialistų trūkumo.

 

2025 m. bendras krovinių, gabenamų Rusijos Arkties jūromis, kiekis, palyginti su ankstesniais metais, sumažėjo 2,3 proc. ir pasiekė 37,02 mln. tonų. SGD sudarė 58 proc. šios sumos, 21 proc. – žalia nafta ir naftos produktai, o 4 proc. – dujų kondensatas.“

 

 

Western Sanctions Are Making China AI Superpower, and Russia - Superpower in Transportation: Russia has found a new route for oil. Shadow fleet heads through the Arctic

Western sanctions reshape global trade by pushing China to advance its domestic artificial intelligence capabilities and enabling Russia to pioneer Arctic shipping corridors, though both face steep economic and security hurdles.

China's AI Growth Under Sanctions

•           Technology bypass: U.S. export controls and chip restrictions have failed to stop Chinese labs from innovating, prompting local breakthroughs and efficient model training techniques. High efficiency makes Chinese models cheap and popular. They are taking over the world. Americans are crying. They don’t know what to do.

•           Retaliation and friction: Beijing condemns U.S. and allied restrictions as "AI hegemony", while Washington continues to expand import bans on connected hardware and robotics.

Russia's Arctic Oil and Shadow Fleet

•           Northern Sea Route: Restricted from traditional European routes, Russian energy exports increasingly travel via the Arctic to Asian markets like China. The Arctic route is getting better as ice is melting in global warming. The Arctic route is much shorter than usual transportation links between Europe and Asia, protected from Muslim insurgency. The West is not crying here yet. It should be crying:

 

“Russia is increasingly utilizing the Northern Sea Route (NSR) to export oil. A record number of "shadow fleet" tankers have gathered in the Arctic, aiming to bypass threats in the Black Sea and hinder Western efforts to control the transport of Russian commodities.

 

In recent months, Ukraine has destroyed or damaged hundreds of vessels linked to Russia. On July 18, Robert Browdy, commander of Ukraine’s Unmanned Systems Forces, reported that 172 vessels had been disabled—118 in the Sea of ​​Azov and 54 in the Black Sea. The majority of these were tankers.

 

The attacks aimed to isolate Crimea and prevent Russia from profiting from energy exports. Additionally, Western nations have been seizing shadow fleet tankers. Consequently, the Kremlin has begun seeking routes that are safer for its business operations.

 

Shadow fleet heads through the Arctic

 

The Northern Sea Route (NSR) is becoming increasingly important in this context. Last year alone, approximately 13.1 million barrels of crude oil were transported via the NSR. Now, the unprecedented volume of cargo prepared for transport at the start of the peak shipping season underscores the Arctic route's growing significance, given the rising geopolitical and military risks Russia faces on more traditional routes, notes the specialized publication *gCaptain*.

 

A flotilla of tankers gathered between the Novaya Zemlya archipelago and the Taymyr Peninsula has already begun preparations to transport approximately 8 million barrels of oil along the Northern Sea Route (NSR). This represents about 60 percent of the total volume of Russian oil exports via the Arctic to Asian countries throughout the shipping season of last summer and autumn.

 

On Monday, August 3, about 15 tankers were sailing along Russia's northern coast or anchored in the Kara Sea, awaiting a nuclear-powered icebreaker or improved ice conditions to continue their journey eastward, according to the specialized publication *gCaptain*.

 

Russia lacks ice-class tankers

 

Three tankers are already sailing the Northern Sea Route (NSR), with one nearing the end of its Arctic voyage. Two others are heading north from the west, via the Norwegian and Barents Seas. Three nuclear-powered icebreakers have been assigned to escort the tankers.

 

Navigation in the Arctic seas typically lasts from late June to early November, with the ice at its thinnest in late September.

 

Tankers can now navigate the NSR independently if they are of the so-called "ice class" (featuring specially reinforced hulls). This is where a major obstacle arises: a shortage of ice-class tankers.

 

Western sanctions have cut the Kremlin off from orders placed with foreign shipyards. Furthermore, the Russians—who, prior to Putin’s war, predominantly ordered such vessels from South Korean shipyards—are unable to build them domestically. Consequently, the construction of vessels specialized for navigating thick ice—forced by sanctions to take place at the Zvezda shipyard in the Far East—is progressing slowly due to issues with equipment supplies (which rely heavily on imports), financing (costs are rising year by year), and a lack of specialists.

 

In 2025, the total volume of cargo transported via the seas of the Russian Arctic fell by 2.3 percent compared to the previous year, reaching 37.02 million tonnes. LNG accounted for 58 percent of this total, crude oil and petroleum products for 21 percent, and gas condensate for 4 percent.”