LesenForschungRadarAnlageframework
Anmelden / Registrieren
AlleTechnikProdukteInvestierenBörseSonstiges

Lesen

2026-10-30115 Beiträge

Aktualisieren

Chinese ADAS Provider Zhuoyu Tech Launches 'Panxing' for Commercial Vehicles, Covering Over 90% of Domestic Heavy Truck Market

Thema · 卓驭科技磐行系统Wesentlich
2026-10-30 12:04 GMT+8

On October 29, Chinese intelligent driving solution provider Zhuoyu Technology officially launched the 'Panxing' advanced driver-assistance system (ADAS) for commercial vehicles, marking an expansion from passenger cars to heavy trucks, buses, and light trucks. The solution comes in two versions: PDS 4, which is already in mass production and supports highway Navigation on Autopilot (NOA) for long-haul logistics, and PDS 5, planned to support urban NOA for city-driving scenarios.

Zhuoyu CEO Shen Shaojie noted that the penetration rate of intelligent driving in China's commercial vehicle sector is currently less than 1%, significantly lower than the over 30% adoption of L2+ systems in passenger cars. He predicts a rapid growth phase for commercial vehicle ADAS in the next two to three years. For suppliers with existing passenger car data, migrating trained large models is more cost-effective than developing independent systems from scratch. Zhuoyu's passenger car ADAS has accumulated over 3 billion kilometers of data, supplemented by approximately 400 million frames of specific heavy-truck data for targeted training.

Technically, Liu Sikang, VP of Zhuoyu's commercial vehicle business, summarized the approach as 'unified foundation, scenario enhancement, and engineering adaptation.' To address challenges like long braking distances and large blind spots in heavy trucks, the new solution features the 'JiMu 2.0' perception system, placing LiDAR inside the cabin to mitigate weather interference and achieving a detection range of up to 500 meters on highways. The system also uses visual recognition to estimate trailer angles and adjusts steering and braking based on real-time load information.

Economic viability is critical for commercial clients. According to Zhuoyu, its fuel-saving algorithm, trained via reinforcement learning, achieved a 4.5%-7.5% reduction in fuel consumption across over 200 tracked mass-produced heavy trucks in the past two months. Assuming 800 km of daily operation, each truck can save approximately 20,000 to 30,000 RMB annually in fuel costs alone. Including reductions in driver fatigue, accidents, and insurance premiums, Zhuoyu estimates the system pays for itself within about one year.

In terms of market reach, Zhuoyu has partnered with six major domestic heavy truck manufacturers: XCMG, Shaanxi Automobile Group, Sinotruk, FAW Jiefang, Foton Motor, and Dongfeng Commercial Vehicles. Zhuoyu states these companies collectively hold over 93% of the domestic heavy truck market. The first mass-production client, Shaanxi Auto, began delivering the Delong X5000 equipped with PDS 4 in June, with hundreds of units operating on Yunda Express lines. XCMG's Hanjing new energy trucks and Sinotruk's Sitrak high-end models are scheduled for delivery by year-end. Additionally, Zhuoyu is advancing overseas collaborations, with European projects expected to launch next year.

Quellen:leiphone.com

Forschung

Alibaba Cloud Launches Apsara Stack International v3.18 for Global Localized Deployment

Thema · 阿里云飞天国际版Wesentlich
2026-10-30 12:18 GMT+8

On October 29, Alibaba Cloud launched the new version of its enterprise-grade private cloud solution, Apsara Stack International v3.18, at the international Apsara Conference. The release provides a trusted AI cloud solution designed to support governments, telecom operators, and large enterprises globally in upgrading their AI infrastructure. Currently, Apsara Stack has achieved localized deployment in more than 20 countries and regions worldwide.

As AI accelerates into critical sectors such as government affairs, finance, and healthcare, overseas institutions face growing demands for autonomous control over data, models, and cloud platforms. The new international version offers over 80 full-stack cloud services, maintaining advantages like 100% local deployment and fully independent operation without internet connectivity. It further integrates proprietary models, private knowledge bases, and Agent capabilities. Through joint optimization of chips, cloud platforms, and models, the solution enhances AI computing efficiency, supported by AI-native operations management and 24/7 global technical support to lower the barrier for self-operation.

In terms of specific implementations, as an Olympic Worldwide Partner, Alibaba Cloud will provide local dedicated cloud services for the 2026 Dakar Youth Olympic Games opening on October 31. Using a dual-availability-zone high-availability architecture, it supports ticketing and related digital operations systems. This marks the first time Apsara Stack's dedicated cloud has been applied to an Olympic event. Additionally, in Sarawak, Malaysia, local digital infrastructure provider SAINS built a government dedicated cloud based on Apsara Stack, deploying a dual-AZ disaster recovery architecture across two data centers in Kuching to upgrade over 300 traditional applications to cloud-native structures.

In the enterprise AI sector, international companies such as AstraZeneca and Boston Scientific have developed exclusive AI applications based on Qwen and Alibaba Cloud's enterprise-level MaaS platform OpenTrek. According to Alibaba Cloud, AstraZeneca applied AI to pharmacovigilance and intelligent business analysis scenarios, improving report writing efficiency by 300%. Boston Scientific integrated approximately 3TB of professional documents to build a medical device knowledge base, increasing knowledge retrieval efficiency by 80%. Both companies achieved 98% accuracy in intelligent business analysis queries and knowledge base Q&A scenarios.

IDC reports indicate that Alibaba Cloud holds the top market share in China's hybrid cloud segment and was rated a 'Leader' in the Asia-Pacific Sovereign AI Platform MarketScape assessment. Leveraging technology and operational capabilities validated through large-scale domestic applications, Alibaba Cloud is bringing mature dedicated cloud technologies to international markets, helping local partners independently build and operate cloud and AI platforms.

Quellen:leiphone.com

Forschung

Korean Survey on AI Customer Service: Expected Job Cuts Did Not Materialize; Clear Human-AI Handoff Standards Prove Key to Success

Wesentlich
2026-10-30 07:44 GMT+8

South Korean HR tech company Remember&Company released its 'AI Customer Experience Report' on October 30, detailing the actual workforce changes and operational outcomes for major Korean enterprises after implementing artificial intelligence in customer service. The report is based on a survey of CS practitioners and decision-makers at key Korean firms.

The data shows that while 36% of companies without AI expected headcount reductions upon adoption, only 18% of those actually operating AI reduced their staff. Conversely, 29% increased their customer service personnel, and 45% opted to reallocate tasks rather than cut jobs. This indicates that the widespread expectation of 'AI inevitably leading to customer service layoffs' has not fully materialized on the ground.

The report also highlights that customers still prefer human interaction even after AI implementation. 50% of companies reported that customers most frequently requested transfers to human agents, and 45% faced cases where customers resubmitted the same inquiries. Although response speeds improved, they did not meet customer expectations for resolution. Furthermore, there is a gap between internal metrics (such as response time and processing cost), which showed a 65% improvement rate, and external customer perception metrics like satisfaction, which showed only a 52% improvement rate.

A critical factor determining the success of AI customer service projects is not just AI performance, but whether clear 'role division standards between AI and human agents' were established beforehand. Among companies whose AI results exceeded expectations, 78% had such standards in place, compared to only 32% among those who fell short. However, currently, only 8% of AI-adopting companies have documented these standards, with most relying on employee experience.

As AI applications expand, skill requirements for customer service staff are also shifting. 53% of companies cited 'ability to use AI tools' and 44% cited 'data interpretation and analysis skills' as becoming more important, surpassing emotional response and empathy capabilities (30%). This reflects a new demand for human judgment to monitor AI and intervene at appropriate times.

Currently, 74% of surveyed companies have adopted or are piloting AI for customer service, but only 6% have expanded it across all channels and the entire organization. 53% plan to increase AI investment within the next year, while only 4% plan to reduce it. New application areas focus primarily on inquiry summarization/auto-classification (30%), voice bots (26%), and real-time assistance for human agents (24%), rather than direct-to-customer chatbots (19%).

Quellen:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Forschung

Former Xiaomi Executive Wang Teng Launches ISHO with AI Smart Ring and Temperature-Control Bed

Thema · ISHO智能睡眠硬件Kurzfassung
2026-10-29 18:47 GMT+8

On October 29, ISHO (Jinri Yixiu), a technology brand founded by Wang Teng, former General Manager of Xiaomi China Marketing, held its inaugural launch event in Beijing. The company officially introduced two AI sleep technology ecosystem products: the ISHO Ring smart ring and the ISHO iFlow intelligent temperature control system.

The ISHO Ring focuses on sleep health monitoring, featuring an integrated design and equipped with the Apollo 330B Plus chip to support on-device AI model execution. The company claims its sensor configuration surpasses leading industry competitor Oura 5 and includes light monitoring capabilities. The ring itself is priced at 1,799 yuan, while the version with a charging case costs 2,199 yuan.

The other product, the iFlow intelligent temperature control system, consists of a main unit and a mattress, dynamically regulating user body temperature via a water circulation system to intervene in sleep quality. The system features the MediaTek Genio 420 computing platform and can locally run the company's self-developed physiological signal large model. The standard version is priced at 9,999 yuan, with the Pro version starting at 12,999 yuan.

During the launch, Wang Teng revealed that the company's future business model will consist of two layers: smart hardware sales and health data service subscriptions. He explicitly stated that ISHO is positioned as a global brand, with overseas markets being a priority. The team currently exceeds 100 people and collaborates with experts from Tsinghua University and Peking University on sleep research.

Quellen:zhidx.com

Forschung

US Crypto Exchange Coinbase Reports $719M Q3 Net Income as Revenue Falls 35% YoY

Thema · 特斯拉交付量Wesentlich
2026-10-29 23:09 GMT+8

U.S.-based cryptocurrency exchange Coinbase Global Inc. released its third-quarter financial results on October 29, 2026, reporting a net income of $719 million, marking its first profit in a year. However, this profit was largely driven by approximately $715 million in unrealized gains on crypto assets held on its balance sheet. Excluding these investment gains, adjusted net income stood at $78 million.

Despite returning to profitability, Coinbase's core operations faced headwinds. Total revenue for the quarter came in at $1.22 billion, flat from the previous quarter but down 35% year-over-year, in line with market expectations. Transaction revenue slipped slightly to $590 million, reflecting lower trading volumes amid multiyear lows in crypto volatility.

Emerging businesses showed strong momentum. Coinbase's prediction markets surpassed $200 million in annualized revenue, growing at an 82% quarterly compound rate since launch, making it one of the fastest-growing products in the company's history. Subscription and services revenue reached $570 million, accounting for nearly half of net revenue, with stablecoins contributing $307 million. Average USDC held in Coinbase products hit a record $22 billion.

On operational efficiency, adjusted EBITDA jumped 60% quarter-over-quarter to $333 million, extending the company's streak of positive adjusted EBITDA to 15 consecutive quarters. Coinbase also lowered its full-year 2026 adjusted expense guidance midpoint by $100 million to between $4.15 billion and $4.3 billion.

Quellen:investor.coinbase.com

Forschung

Chinese Optical Giants Slump Despite Strong Q3 Earnings as Growth Slowdown Fears Mount

Thema · 中国光模块估值争议Wesentlich
2026-10-30 12:27 GMT+8

Shares of major Chinese optical communications suppliers extended a two-day sell-off on Friday, despite upbeat third-quarter earnings driven by the global artificial intelligence infrastructure boom. During morning trading in Shenzhen, Chengdu-based Eoptolink Technology dropped as much as 15 per cent, while Suzhou TFC Communication slid 10.3 per cent and industry leader Zhongji Innolight fell over 7 per cent. By the midday break, the trio had trimmed their losses: Eoptolink was down around 9 per cent, TFC pared its decline to 4 per cent, and Innolight also steadied. The retreat followed steep drops on Thursday, when TFC plunged 19.8 per cent, Eoptolink tumbled 11.1 per cent and Innolight fell 8.5 per cent.

The market rout came even as the firms reported robust headline earnings growth. Innolight on Thursday posted a 173.7 per cent surge in third-quarter net profit to 8.59 billion yuan (US$1.29 billion), as revenue jumped 171.8 per cent to 27.77 billion yuan. The company credited AI infrastructure investment and rising sales of high-end 800-gigabit and 1.6-terabit optical modules for driving top-line growth in the first nine months. Beneath the headline expansions, however, sequential momentum showed signs of cooling, sparking valuation worries among investors.

Eoptolink on Thursday reported quarterly net profit of 3.84 billion yuan, up 61.2 per cent year on year, on an 80.3 per cent rise in revenue to 10.94 billion yuan. But on a quarter-on-quarter basis, profit dropped 19.1 per cent and revenue shrank 13 per cent, according to calculations based on company filings. Cash flow also raised concerns. Eoptolink’s nine-month operating cash flow contracted 68.4 per cent to 1.46 billion yuan, hit by increased materials procurement outlays and tax payments. The third quarter saw an operating cash outflow of about 152 million yuan.

Meanwhile, Suzhou-based TFC earlier this month posted quarterly net profit of 879.2 million yuan, up 55.3 per cent, alongside a 31.5 per cent revenue increase to 1.92 billion yuan. The figures fell short of market expectations. Citigroup said TFC’s revenue missed its own forecast by 34 per cent and lagged Bloomberg consensus estimates by 24 per cent. While gross margins were better than expected at 65 per cent, Citi cut its 2026 and 2027 earnings forecasts by 19 per cent each, citing slower revenue recognition and more cautious assumptions about the ramp-up of optical engines.

Industry analysts and company executives, however, maintained that downstream appetite remained intact. “The faster revenue growth in 3Q26 implies that the component shortage is gradually easing,” Morgan Stanley analysts wrote in an October 28 research note. The companies have moved to reassure jittery investors. Innolight said in a disclosure last week that its order book was full, with major overseas clients placing orders through 2027, while offering clear demand guidance into 2028. Eoptolink said mass shipments of its 1.6T modules had entered a sustained shipment growth phase, expressing confidence in the company’s business prospects this year and next.

Quellen:South China Morning Post

Forschung

TE Connectivity Japan introduces humanoid robot trial at Kakegawa factory

Kurzfassung
2026-10-30 20:00 GMT+8

According to EE Times Japan, TE Connectivity Japan, a subsidiary of connector and sensor manufacturer TE Connectivity, has begun a trial introduction of a humanoid robot at its Kakegawa factory in Shizuoka Prefecture since August 2026. This marks a concrete attempt to apply "physical AI" technology in a real-world manufacturing setting.

Physical AI refers to technology that uses sensors to acquire information from the real world, allows AI to recognize and make judgments, and reflects the results in the actions of robots or equipment. Despite the emergence of high-performance AI models, manufacturing sites face complex real-world factors such as positional deviations, part tolerances, and contact forces, making it difficult to achieve practical use with software alone. This trial aims to verify the adaptability of AI robots in an actual industrial environment.

Quellen:EE Times Japan

Forschung

OCEANO HK HOLDING Releases Open Weights for Hi3D V1.0, Enabling Local High-Resolution 3D Shape Generation

Kurzfassung
2026-10-30 11:14 GMT+8

Hong Kong-based company OCEANO HK HOLDING released the open weights and inference code for its AI 3D shape generation model, Hi3D V1.0, on October 23. Based on the architecture formerly known as Sparc3D, the model takes a coarse mesh and a reference image of the same object as input, outputting high-resolution shape data without textures or materials.

Hi3D V1.0 supports a maximum resolution of 1,536³, executing one to three reconstruction passes depending on the required detail. The training dataset was expanded from approximately 500,000 to about 2 million curated 3D assets. According to the vendor's self-testing, the model achieved the highest scores in multi-view consistency and detail richness, reaching a 90% slice success rate.

Running the model requires an NVIDIA GPU with Compute Capability 8.0 or higher, CUDA Toolkit 12.0 to 12.5, and a Linux environment. Operational verification has currently been conducted only on an NVIDIA H20 with 96GB of memory; smaller capacities remain unverified.

The accompanying 'Hi3D Open License Agreement Version 1.0' permits royalty-free commercial use. However, if a product or service integrating the model exceeds 200,000 monthly active users or $1 million in monthly revenue in the most recent calendar month, the developer must notify the company within 60 days and apply for a separate written license. Commercial products must also display a 'Powered by Hi3D 1.0' label.

Quellen:PC Watch

Forschung

GitHub COO Kyle Daigle: Agent Traffic Patterns Upend Scaling Assumptions, Code Verification Becomes the Next Bottleneck

Thema · GitHubWesentlich
2026-10-29 21:30 GMT+8

At the GitHub Universe 2026 conference held in San Francisco, GitHub Chief Operating Officer Kyle Daigle and Microsoft CoreAI Executive Vice President Jay Parikh (now leading GitHub) faced a sharp question: with the competitive platform Entire, founded by former CEO Thomas Dohmke, just launching, does GitHub need to worry about being replaced? Daigle’s answer was: the world does not need a new GitHub; it needs GitHub to evolve for the Agent era. However, the technical rationale supporting this judgment is worth examining more closely than the slogan itself.

Daigle explained that GitHub began experiencing a series of unexpected failures starting this summer—services designed for human developer traffic patterns, which had run stably for a long time, suddenly crashed. He attributed this not to insufficient capacity, but to fundamental differences in behavior patterns: Agents operate in ways completely different from humans. Some services scaled well under human usage but exhibited failure points the team had not anticipated under concurrent Agent access. He explicitly stated that simply increasing core counts and horizontal scaling is no longer sufficient; core services themselves must be rewritten because the behavioral differences are too great.

Parikh added that the current growth rate makes incremental optimization meaningless. With extensive experience scaling infrastructure at Facebook and Akamai, he judged that achieving five to ten percent improvements now only buys one or two weeks. Therefore, GitHub decided to completely rewrite its core services rather than merely adding capacity.

The specific architectural responses include two levels. First, separating compute from storage within the Git infrastructure allows both to scale independently according to their respective needs. Parikh explained that compute and storage load variation patterns differ based on traffic type, time of day, user identity, and product form; separation enables independent optimization. Second, migrating from GitHub’s own data centers to Azure provides elastic capacity, geographic flexibility, and redundancy—features fixed facilities cannot offer. As of early October, the migration was approximately 60% complete.

However, Daigle believes the true new constraint lies not in generation, but in verification. When five or ten Agents simultaneously submit code to a repository, running full builds for every submission is neither desirable nor necessary. He proposed several possible directions for response: signing locally executed builds and submitting them alongside the Agent’s work, selecting smaller sets of relevant tests, and exploring proof-driven development methods. He emphasized that the correct combination of testing, evaluation, and real-time verification depends on the specific application scenario—automakers will not let Agents autonomously check whether brake calipers are engaged.

In terms of competitive narrative, Parikh positioned GitHub as a platform company: providing a complete workflow while also offering primitives that can be used independently, allowing enterprises to retain choice. He explicitly stated that GitHub has both products and platforms, whereas much of the current competition consists of single-product tools. Daigle analogized the current moment to the end of the zero-interest-rate era—when numerous sophisticated but extremely narrow developer tools emerged, and ultimately many teams joined GitHub. This analogy implies that the Agent era will similarly experience tool fragmentation, but platform integration remains the final direction.

Quellen:The New Stack

Forschung

OpenAI CEO Sam Altman Says AI Is Opening an Unprecedented Entrepreneurial Window, With Potential to Spawn Companies Larger Than OpenAI

Wesentlich
2026-10-29 20:30 GMT+8

In an interview with the Stanford Graduate School of Business, OpenAI CEO Sam Altman expressed 'persistent disappointment' toward Gen Z students. His criticism was specific: these students would rather do what was popular five years ago than attempt new paths. In his view, now is precisely the best time to take those new paths.

Altman's argument proceeds as follows: AI will indeed make many tasks currently performed by humans better, cheaper, and faster, and it will fully take over certain large segments of current work. However, this disruption creates opportunities for ambitious workers—provided they are willing to consider which new business models can be developed leveraging AI.

His most striking assessment is that the current starting point is 'unprecedentedly open.' AI has opened a rare time window for entrepreneurs to build areas that were difficult or even impossible just a few years ago. Based on this technology, companies could emerge that are far larger and more significant than OpenAI itself.

Altman's advice to the Stanford students was: If you enter the world to start working now, your careers will likely begin during a period of the greatest change. Those with great potential, drive, ideas, new perspectives, and new skills tend to thrive particularly well 'when the ground is shaking the most.'

It should be noted that these assessments come from public statements made by Altman himself, reported by Fortune and cited by t3n. As the leader of OpenAI, framing AI entrepreneurial opportunities as 'open' may serve a commercial narrative positioning OpenAI as an infrastructure platform. The assertion that companies 'larger than OpenAI' will emerge lacks support from specific cases, pathways, or timelines; it represents a forward-looking viewpoint rather than a verifiable fact.

Quellen:t3n.de

Forschung

US Aviation Software Firm Surf Air Mobility Signs Five OperatorOS Deals, Expands Palantir Partnership

Kurzfassung
2026-10-29 18:30 GMT+8

Los Angeles-based air mobility platform Surf Air Mobility Inc. (NYSE: SRFM) issued a business update on October 29, 2026, highlighting commercial progress for SurfOS, its AI-enabled software operating system built on technology from Palantir Technologies (NASDAQ: PLTR). The company stated it has signed definitive agreements for OperatorOS with five small and mid-sized Part 135 operators (a US Federal Aviation Administration category for on-demand aviation): Sprintbach Aviation, SkyDance Air, Air Fuga, Clipper Aviation, and Flying Zebra. Under these deals, Surf Air Mobility will earn a percentage of revenue for all trips managed through the system.

The update also emphasized an expansion of the partnership with Palantir. Under an exclusive teaming agreement, Palantir will co-sell SurfOS to enterprise prospects, with its forward-deployed engineers supporting sales and implementation. Additionally, the company appointed former Palantir executive Barrett Brown as President of SurfOS to lead go-to-market strategy and commercial expansion.

Beyond the five new operator contracts, the company confirmed a multi-year Enterprise BrokerOS agreement with premium charter service provider Wheels Up, valued at up to $12 million over the term. BrokerOS has been commercially live since December 2025, with internal data showing it helps increase bookings and shorten quote-to-close times. Management noted an active enterprise pipeline involving large operators, brokerages, and aircraft manufacturers that they believe could be worth tens of millions of dollars in annual revenue, though this figure remains a company estimate rather than confirmed income.

Quellen:Yahoo Finance

Forschung

US Air Force Seeks AI to Auto-Detect Mid-Air Refueling Errors via SBIR Solicitation

Kurzfassung
2026-10-30 09:10 GMT+8

According to TechRadar, the US Air Force is exploring the use of artificial intelligence to identify mistakes during aerial refueling operations and improve how crews receive performance feedback. The project falls under the Small Business Innovation Research (SBIR) program, which invites companies to develop technological solutions addressing specific military requirements. Interested developers have until November 25, 2026, to submit applications outlining how their proposed systems could satisfy the service's operational and training requirements.

The proposed system would analyze footage captured by cameras on the KC-46 Pegasus, a refueling aircraft equipped with cameras supporting remotely controlled fuel transfer operations. Its remote vision equipment allows boom operators to guide the refueling apparatus using camera feeds and 3D displays instead of direct observation. However, the aircraft's existing viewing technology has experienced persistent technical difficulties.

Using computer vision and machine learning, the technology would examine aircraft movements and determine whether crews maintained appropriate separation and alignment. It would also monitor the refueling boom, comparing its movements against aircraft trajectories to identify potentially dangerous deviations during fuel delivery. Measurements would include aircraft speed, relative positioning, approach velocity, and alignment, allowing instructors to examine specific maneuvers rather than relying on general observations.

Currently, assessments depend substantially on instructor observations, crew evaluations, and manually examining recorded footage, according to the military's research solicitation. These methods can produce inconsistent judgments, particularly when instructors interpret the same maneuver differently or overlook brief errors during complicated operations. Therefore, the software’s assessment will include graphical overlays, color-coded performance maps, and chronological displays identifying significant events during individual refueling exercises.

Quellen:TechRadar

Forschung
Weiter →
Anmelden / Registrieren
LesenForschungRadarAnlageframework

It should be noted that all of the above represents self-assessment and strategic narrative from GitHub’s leadership. Dohmke’s response directly questioned the premise: entrusting the future of software development to a single large tech company is itself a problem. Daigle’s description of Agent behavioral patterns causing unexpected failures is an attributive statement; while GitHub has published a public post-mortem of the August outage, this article did not read the specific technical details of that report. The argument regarding the verification bottleneck remains a directional judgment at present, with no deployed solutions yet available to prove its feasibility.