<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>DATA TIMES (English)</title><link>https://datatimes.jp/en</link><description>The world is more interesting through data.</description><language>en</language><item><title>Can a robot learn a craftsman&apos;s &quot;why&quot;? ATTRAC LAB&apos;s Tomoyuki Izu on where Japan can win in physical AI</title><link>https://datatimes.jp/en/articles/attraclab-robot-brain-interview</link><guid>https://datatimes.jp/en/articles/attraclab-robot-brain-interview</guid><pubDate>Wed, 21 Oct 2026 23:00:22 GMT</pubDate><description>AI that moves robot bodies, often called physical AI, is advancing fast; humanoid robots have improved dramatically in a single year. So where can Japan compete? Tomoyuki Izu, who has founded two drone companies and now runs the robot developer ATTRAC LAB near Tokyo, points to the robot&apos;s brain. His idea: have veteran workers talk through what they are doing, and train the AI on those words along with their movements, so robots inherit not just the motion but the reasons behind it. We asked how it works, where it starts (drones and construction sites), and the rural future he has in mind.</description></item><item><title>To double its shipbuilding, Japan is 60,000 workers short. Can wall-climbing robots and AI that learns from master craftsmen close the gap?</title><link>https://datatimes.jp/en/articles/shipbuilding-robots-ai-japan</link><guid>https://datatimes.jp/en/articles/shipbuilding-robots-ai-japan</guid><pubDate>Tue, 20 Oct 2026 23:00:21 GMT</pubDate><description>Car plants are full of robots, yet ships are still mostly built by hand. Japan has now decided to build about twice as many ships by 2035, and at today&apos;s productivity it would be about 60,000 workers short. DATA TIMES collected 72 robot and AI projects at Japanese shipyards and 28 shipyard tech job postings to show why ships are so hard for robots, how many robots are actually running, and how far Korea, China and the US are ahead, from wall-climbing quadrupeds to AI that learns from master craftsmen.</description></item><item><title>In Korea, foreign tourists can’t even hail a taxi by app. CES-winning CrossHub wants your passport to be your wallet</title><link>https://datatimes.jp/en/articles/crosshub-travel-pass-interview</link><guid>https://datatimes.jp/en/articles/crosshub-travel-pass-interview</guid><pubDate>Mon, 19 Oct 2026 23:00:20 GMT</pubDate><description>In South Korea, foreign visitors struggle to hail a taxi by app, order food delivery or book KTX high-speed rail tickets online, because those services sit behind residents-only identity checks. Seoul startup CrossHub, a CES 2026 Best of Innovation winner in fintech, is building Travel Pass: a card and app that verifies travelers with their passport before the trip and lets them use local services like residents do. We met Ra Sangyong, head of strategic initiatives, to talk about zero-knowledge identity checks, money that only works in one city, student IDs that keep graduates in town, and why Japan comes next.</description></item><item><title>Only 3% require industry knowledge. We read 326 Japanese job postings for engineers who take AI into the field, and the real ask is to see it through until people use it</title><link>https://datatimes.jp/en/articles/fde-jobs-japan-2026</link><guid>https://datatimes.jp/en/articles/fde-jobs-japan-2026</guid><pubDate>Sun, 18 Oct 2026 23:00:01 GMT</pubDate><description>Forward deployed engineers embed with customers, or with their own company&apos;s departments, to make AI work in real operations. DATA TIMES read the full text of 326 such job postings in Japan from 168 companies and recorded what each one asks for. Seven in ten hand one person the whole job: discovery and requirements, prototype to production, and operations and adoption. Only 3% require industry knowledge. Here is what Japanese employers actually want from the engineers who take AI into the field.</description></item><item><title>280 data leaks in 47 days: counting every one in Japan reveals apps&apos; back doors under attack, and the arrival of AI that can hack</title><link>https://datatimes.jp/en/articles/data-leaks-hubs-2026</link><guid>https://datatimes.jp/en/articles/data-leaks-hubs-2026</guid><pubDate>Sat, 17 Oct 2026 23:00:01 GMT</pubDate><description>DATA TIMES collected every leak and intrusion disclosed in Japan from September 1 to October 17, 2026: 280 in all. The main techniques were taking apart smartphone apps to abuse weakly protected APIs, and breaking into contractors and cloud platforms where data gathers. Just before the wave, an open-weight model with strong hacking skills became downloadable by anyone, as experts had long warned. Do defenders need AI too, and how do South Korea and the EU punish breaches?</description></item><item><title>A disaster register for 165,411 buildings, from free data alone: we had an AI layer satellite, terrain, quake and land-price data onto Fujisawa&apos;s PLATEAU model</title><link>https://datatimes.jp/en/articles/plateau-fujisawa-165000-buildings</link><guid>https://datatimes.jp/en/articles/plateau-fujisawa-165000-buildings</guid><pubDate>Fri, 16 Oct 2026 23:00:48 GMT</pubDate><description>We asked an AI to pull all 165,411 buildings in Fujisawa, Japan, from the PLATEAU 3D city model and write five free public datasets onto each one: GSI landforms, large-scale fill sites, J-SHIS earthquake shaking odds, Sentinel-2 satellite change and official land prices. In about an hour and a half it had a city-wide disaster register. Among the tenth of buildings most prone to strong shaking, only one in four would be colored on a flood hazard map; satellites found a development site before the 3D city did; and tsunami-zone land still rose in price. Here is how we built it, where we tripped up, and what a register can and can&apos;t say about one building, with reproduction steps.</description></item><item><title>Debt up 22x in three and a half years: what the &apos;CoreWeave shock&apos; reveals about how AI&apos;s compute factories get built</title><link>https://datatimes.jp/en/articles/coreweave-shock-ai-debt</link><guid>https://datatimes.jp/en/articles/coreweave-shock-ai-debt</guid><pubDate>Thu, 15 Oct 2026 23:00:34 GMT</pubDate><description>On October 15, Japan&apos;s Nikkei 225 fell 3,318 yen, its third-largest point drop ever, with Japanese media calling it the &apos;CoreWeave shock&apos;. CoreWeave, which rents GPU computing to AI companies, has grown its debt about 22-fold in three and a half years to $35.6 billion, and in the first half of 2026 it paid more in debt service than it earned from operations. Using its SEC filings, we trace how AI compute factories are financed by borrowing against GPUs and customer contracts, what worries lenders, and what holds the structure up.</description></item><item><title>Start with AI, not with IT. Korea&apos;s realizable wants to bring AI to factories one industrial park at a time</title><link>https://datatimes.jp/en/articles/realizable-ai-factory-manager-interview</link><guid>https://datatimes.jp/en/articles/realizable-ai-factory-manager-interview</guid><pubDate>Tue, 13 Oct 2026 23:00:49 GMT</pubDate><description>The usual path to factory digitalization goes systems first, then data, and AI last. Most small plants run out of time and money long before they get there. realizable, a start-up based in Incheon and Seoul, flips the order. Its &quot;AI Factory Manager&quot; takes the name of a single product, proposes the processes and the numbers worth watching, and can build a working dashboard in about 30 minutes. And instead of piloting one plant at a time, it tests 10 or 20 plants at once through an industrial park. We spoke with CEO Dongmyoung Won about how to connect pen-and-paper shop floors to AI, and what he wants to offer Japan.</description></item><item><title>Our editor wants to open a seaside café. An AI wired into Japan&apos;s government data skipped the busiest station — and worked out the rent he can afford</title><link>https://datatimes.jp/en/articles/mlit-mcp-ai-store-location</link><guid>https://datatimes.jp/en/articles/mlit-mcp-ai-store-location</guid><pubDate>Mon, 12 Oct 2026 23:01:48 GMT</pubDate><description>DATA TIMES&apos;s editor-in-chief has a quiet dream: a small café near the sea. So he asked an AI agent connected to Japan&apos;s transport ministry data where to open it, in three casual messages. The AI compared three stations on the Shonan coast in just over three minutes, reversed its own pick on the second message, and on the third built a calculator for sales and the rent he could afford. We show the real screens, check the numbers against official statistics, and lay out what government data still can&apos;t tell you.</description></item><item><title>An AI wrote 719 math papers in two weeks. Did OpenAI just solve the Riemann hypothesis? We counted all 372 results</title><link>https://datatimes.jp/en/articles/openai-math-719-riemann</link><guid>https://datatimes.jp/en/articles/openai-math-719-riemann</guid><pubDate>Sat, 10 Oct 2026 23:01:19 GMT</pubDate><description>OpenAI has released 719 math papers written by an unreleased internal AI model, headlined by a claimed proof of the &quot;quasi-Riemann hypothesis.&quot; So has the 160-year-old Riemann hypothesis finally fallen? No, though the claim could be a big step toward it. We recounted all 372 results ourselves: 36% have a main result checked by computer, about one in five overturns a conjecture, and the manuscripts run to more than 34,000 pages.</description></item><item><title>One extra molecule, and in three rounds it takes over: the self-copying reaction behind the 2026 Nobel Prize in Chemistry</title><link>https://datatimes.jp/en/articles/nobel-chemistry-2026-soai-autocatalysis</link><guid>https://datatimes.jp/en/articles/nobel-chemistry-2026-soai-autocatalysis</guid><pubDate>Fri, 09 Oct 2026 21:44:48 GMT</pubDate><description>Start with 1,000,001 right-handed molecules and 1,000,000 left-handed ones. Three rounds of one reaction later, the right-handed kind makes up more than 99.7% of the mix. Japanese chemist Kenso Soai built that reaction, and it has just earned him the 2026 Nobel Prize in Chemistry, shared with France&apos;s Henri B. Kagan. Here is who discovered what, why your body uses only one &quot;hand&quot; of its molecules, and how the same chemistry shapes the medicines you take.</description></item><item><title>Astroscale isn&apos;t really a space junk removal company. Inside its plan to build roadside assistance for orbit</title><link>https://datatimes.jp/en/articles/astroscale-on-orbit-servicing-interview</link><guid>https://datatimes.jp/en/articles/astroscale-on-orbit-servicing-interview</guid><pubDate>Thu, 08 Oct 2026 23:01:17 GMT</pubDate><description>Satellites and space junk keep piling up around the Earth, and a satellite that runs out of fuel has usually just been left where it is. Japan&apos;s Astroscale wants to build the aftercare that space never had. Astroscale Japan&apos;s PR lead, Satoshi Ito, explains that the real business is roadside assistance for orbit — getting close to satellites to inspect them, refuel them, extend their lives and, finally, clear them away. We look at how its ADRAS-J spacecraft got 15 meters behind an uncooperative piece of debris, why so many of its engineers come from outside the space industry, and what the data says about an increasingly crowded orbit.</description></item><item><title>Anyone can make 3D now. So why don’t job sites use it? We asked the maker of a Gaussian splatting viewer</title><link>https://datatimes.jp/en/articles/wan-machica-3dgs-interview</link><guid>https://datatimes.jp/en/articles/wan-machica-3dgs-interview</guid><pubDate>Thu, 08 Oct 2026 05:05:06 GMT</pubDate><description>With 3D Gaussian Splatting (3DGS), a phone video becomes a photo-real 3D copy of a place. Three years after the paper, it is now part of the standard 3D file format. Yet very few construction companies actually use it for work. We asked Ayumu Mitani, CEO of World Area Networks (WAN) in Hiroshima and maker of the 3DGS viewer Machica, what is missing: how splats differ from point clouds, how to overlay design data, why his team lets users steer the product, and why you can now shoot a scene again after you have left it. Part 1 of our series on WAN.</description></item><item><title>40,000 papers a month, double in two years: we counted every arXiv paper to see why it now caps submitters at two</title><link>https://datatimes.jp/en/articles/arxiv-submission-limit</link><guid>https://datatimes.jp/en/articles/arxiv-submission-limit</guid><pubDate>Wed, 07 Oct 2026 02:09:09 GMT</pubDate><description>arXiv, the free preprint server where much of the world&apos;s research appears first, began limiting each submitter to two new papers a month on October 1, 2026. September brought a record 40,363 submissions, almost twice as many as two years earlier. Is AI to blame? We recounted all 38,908 papers posted that month and found a surge in math, a 2.3x jump in solo-authored papers, and a cap that would block only a few percent of papers.</description></item><item><title>Ten days, three Korean cities: after the tourists, Korea wants people to work with</title><link>https://datatimes.jp/en/articles/korea-tourists-young-exodus</link><guid>https://datatimes.jp/en/articles/korea-tourists-young-exodus</guid><pubDate>Tue, 06 Oct 2026 01:39:54 GMT</pubDate><description>Korea keeps coming up whenever we write about AI. So when DATA TIMES&apos;s editor-in-chief got a last-minute invitation to Korea Nomad Gateway, a 10-day program through Incheon, Busan and Jeju, he said yes. This is the diary of working by the sea and sharing tables with local officials, founders and participants from around the world, paired with a data check of what we heard on the road: record tourists, young people leaving, and regions now inviting the world to come and work. The trip ended with partnerships with Korean companies starting to take shape.</description></item><item><title>An AI factory next to a power plant: what Chiba&apos;s ¥2.3 trillion data center is really buying is time</title><link>https://datatimes.jp/en/articles/jera-chiba-ai-datacenter</link><guid>https://datatimes.jp/en/articles/jera-chiba-ai-datacenter</guid><pubDate>Sun, 04 Oct 2026 21:49:45 GMT</pubDate><description>Japan&apos;s largest power producer, JERA, has teamed up with Dell and RHAELM to plan an AI data center of up to 400 MW, worth more than $15 billion, inside the grounds of its Chiba power plant near Tokyo. How big is 400 MW, and how many GPUs could it run? Why build at a power station instead of a normal site? We look at the eight-to-ten-year wait for grid power in Tokyo, Japan&apos;s shifting demand forecasts, the US precedents and the CO2 math.</description></item><item><title>They kept your license photo for 7 years after you quit. What Japan&apos;s Times Car breach teaches about ID checks without images</title><link>https://datatimes.jp/en/articles/times-car-license-image-leak</link><guid>https://datatimes.jp/en/articles/times-car-license-image-leak</guid><pubDate>Sat, 03 Oct 2026 01:19:00 GMT</pubDate><description>A cyberattack on Times Car, Japan&apos;s largest car-sharing service, exposed images of driver&apos;s licenses and other ID documents for about 1.6 million accounts, including former members whose images were kept for seven years. As generative AI makes forgery and attacks more sophisticated, a driver&apos;s license has become a key that also unlocks financial services. We look at how Japan is moving to chip-based ID checks, and at ways to verify identity without holding images at all.</description></item><item><title>Japan spent ¥16 billion putting its cities in 3D. Did it pay off? PLATEAU&apos;s report card</title><link>https://datatimes.jp/en/articles/plateau-report-card-budget-renewal</link><guid>https://datatimes.jp/en/articles/plateau-report-card-budget-renewal</guid><pubDate>Thu, 01 Oct 2026 23:08:58 GMT</pubDate><description>Since 2020, Japan has poured over ¥16 billion (about $107 million) into PLATEAU, a free 3D model of 329 cities. We look at what it was for, which projects delivered and which fizzled, how it compares with the UK, the Netherlands, Germany and Singapore, and what our own count of 306 municipalities reveals: most early 3D cities have never been updated.</description></item><item><title>An AI twin that works while you sleep, and a Pro plan cut in half. Inside OpenAI DevDay 2026, plus a build-off against Claude Opus 5.5</title><link>https://datatimes.jp/en/articles/openai-devday-2026-explained</link><guid>https://datatimes.jp/en/articles/openai-devday-2026-explained</guid><pubDate>Thu, 01 Oct 2026 12:00:00 GMT</pubDate><description>OpenAI&apos;s DevDay 2026 left social media buzzing about dots, an always-on AI &quot;second self,&quot; and fuming that the $200 ChatGPT Pro plan now buys half as much. We sort out what was announced and how people reacted, then hand OpenAI&apos;s GPT-6.1 Sol and Astra the exact same prompts as Claude Opus 5.5 and Fable 5.1 (a space shooter and a rainy 3D Tokyo) to see how far apart they really are. Finally: did Pro actually get more expensive, and who comes out ahead?</description></item><item><title>AI checks crops plant by plant: a former Nikon engineer on automating crop research</title><link>https://datatimes.jp/en/articles/chloros-swalo-image-analysis-interview-v2</link><guid>https://datatimes.jp/en/articles/chloros-swalo-image-analysis-interview-v2</guid><pubDate>Wed, 30 Sep 2026 07:00:00 GMT</pubDate><description>Yuichi Ito saw deep learning&apos;s 2012 breakthrough up close as a Nikon image-analysis engineer. Now CTO of Chloros Inc., he explains how its drone-and-AI system SWALO automates the plant-by-plant visual checks that variety and pesticide trials have long relied on.</description></item><item><title>We plugged AI into Japan&apos;s 3D city model and X-rayed 1,022 buildings near Tokyo Station</title><link>https://datatimes.jp/en/articles/plateau-3d-city-model-hands-on</link><guid>https://datatimes.jp/en/articles/plateau-3d-city-model-hands-on</guid><pubDate>Tue, 29 Sep 2026 11:00:00 GMT</pubDate><description>Japan&apos;s national 3D city model, PLATEAU, now talks to AI. We connected through its MCP server, tallied flood risk for 1,022 buildings near Tokyo Station, and found a medical facility facing a projected 4.1 m flood. Plus: AI that builds a 3D city in hours.</description></item><item><title>AI is starting to beat human typhoon forecasters. That circle isn&apos;t the storm&apos;s size</title><link>https://datatimes.jp/en/articles/typhoon-okinawa-forecast-data</link><guid>https://datatimes.jp/en/articles/typhoon-okinawa-forecast-data</guid><pubDate>Tue, 29 Sep 2026 11:00:00 GMT</pubDate><description>AI has started to beat official US forecasts of tropical storm tracks. We look at where 1,000-scenario AI forecasting shines and struggles, what Japan&apos;s weather agency is doing, and how to read the forecast circle, guided by Typhoon No. 26 and 75 years of data.</description></item><item><title>AI is running short of power and text. Its next meal: your company&apos;s records</title><link>https://datatimes.jp/en/articles/ai-hunger-power-data-context</link><guid>https://datatimes.jp/en/articles/ai-hunger-power-data-context</guid><pubDate>Tue, 29 Sep 2026 11:00:00 GMT</pubDate><description>By 2030, data centers are expected to use nearly as much electricity as Japan generates in a year, and public text for AI training is running low. Tracing AI&apos;s three hungers (power, text and context), we show why its next race is over company records.</description></item><item><title>After AI chat comes the AI canvas. We put Lovart, a $2B+ design AI, to work</title><link>https://datatimes.jp/en/articles/lovart-design-agent-hands-on</link><guid>https://datatimes.jp/en/articles/lovart-design-agent-hands-on</guid><pubDate>Tue, 29 Sep 2026 11:00:00 GMT</pubDate><description>An ad with no film shoot, a VTuber character sheet, a movie-style poster: we made them all with Lovart, a design AI that hit 10 million users in five months. Here&apos;s what impressed us, plus a report from its official hands-on workshop in Japan.</description></item></channel></rss>