News

On December 19, OpenAI released a tweet today, December 19, announcing that iOS, Android and the web-based AI chat robot ChatGPT, which is on-line from now on, is now being transferred to users worldwide. ChatGPT previously had a smart tag that automatically generated a title based on the content of the dialogue, but in the actual use of the scenes, as the number of conversations accumulates, the old and important conversations can easily be flooded with endless new streams of dialogue. User..

ACCORDING TO THE JAPANESE-CHINESE-LANGUAGE NETWORK, THE JAPANESE PRODUCTION FIRM WILL INTRODUCE A HUMAN ROBOT IN ITS OWN FACTORY BY 2027, CARRYING A SELF-STUDY AI TO REPLACE HUMANS IN COMPLEX TASKS SUCH AS WIRING OF ELECTRONIC PRODUCTS. THE JAPANESE-DEVELOPED HUMAN ROBOT HAS TWO ARMS, A COMPUTER AND A CAMERA THAT DRIVES AI, AND TWO FINGERS EACH ON BOTH ARMS; THE ROBOT HAS NO LEGS AND IS EQUIPPED WITH A WHEELED TURNER。

ON THE 12TH OF THIS MONTH, THE LOCAL TIME OF THE U.S.-KNOWN OLD U.S. HIGH SCHOOL OF PUFU DYNASTY ANNOUNCED THAT, STARTING WITH THE NEW STUDENTS WHO WERE ENROLLED IN THE FALL OF 2026, THE GRADUATION REQUIREMENTS FOR GRADUATES FROM INDIANAPOLIS AND THE WEST LAFAYETTE MAIN CAMPUS HAD BEEN ADDED TO THE “AI WORK CAPABILITY STANDARD”. THE PUDU UNIVERSITY BELIEVES THAT STUDENTS AT THE UNIVERSITY MUST HAVE CRITICAL THINKING SKILLS FOR AI, WHICH ARE NEEDED IN THE CURRENT LABOUR MARKET TO UNDERSTAND, ASSESS, USE AI EFFECTIVELY AND KEEP UP WITH FUTURE CHANGES。

Commercial soup technology released yesterday a video-marketing smart body for the electrician business scene, combining four electricians, Agent, and digital anchors, to form the Five Smart Matrix, covering the powerer's operating end-to-end full-link. Commercial soup claims that the half-day event of the real person can be reduced to 10 minutes in the complex of store operations, with an average of 20 times the impact of a single job; the time spent running a double disk of equal quality by live operation of Agent is reduced from the original 2 hours to 20 minutes; the flow of Agent is 5 times the effectiveness of a monitoring list; live site control A..

When asked about the overall performance of the soybean model this year, during the FORCE power conference, the director of the volcano engine Tan waits to say that the overall performance is still good, whether at home or not, but if we look at the world, whether OpenAI or Gemini, it's really good, so we have to work harder here, and certainly our Seedance and Seedream are doing well around the globe. Turning to the 2026 competition in the big model market, Tan would say, first of all, not the competition, but the most important..

Since yesterday, App has had full access to the Gothic Map capability to reach depths around navigation and life services: Users in Applep have integrated routing, surrounding catering and entertainment search, and "routing" shop screening to improve the responsiveness to complex travel needs. This access covers key events such as travel of new energy vehicles, travel restrictions and service information. After the user provides information on the place of departure, destination and renewal of the vehicle, App will automatically plan the charging station along the Gothic route with reference to the data; at the same time, support the production of a compliance travel programme based on the end of the licence plate, the time of departure and the destination, and will plan the route..

AI Agent application Manus recently reached the "0 to $100 million ARR" commercial milestone, with the co-founder Zhang Tao (Community ID "hiidecloud") revealing that commercialization has been in place for less than nine months since the end of March, with an annualized run rate of $126 million. The Manus team's product route is moved from "AI Browser" to the "Agent cloud-operated browser and data access" framework, with the core conviction being "Less history, more information..

Elon Musk, the global leader, is both Chief Executive Officer of Tesla and founder of XAI, both of which are now advancing the AI project. And he seemed optimistic about the future of xAI. According to recent media reports from several sources, during last week’s plenary staff meeting at XAI’s headquarters in San Francisco, Maske threatened to defeat rivals as long as the company could survive the next two to three years. He added that the company's ability to rapidly expand its computing and data capacity would be called "super smart" (i.e...

On 18 December, the Volcanic Engine officially launched the Bean Packed Large Model 1.8 and the Audio/Video Creative Model Feedance 1.5 pro at the ForCE Conference. Authoritative assessment data show that the Bean Packer model is the first in the world in terms of multimodular understanding, generating capacity and Agent capability. According to Tan, the CEO of the Volcanic Engine, by December of this year, the average daily token use of the soybean bun model had exceeded 50 trillion, more than 10 times the same period last year; there were now over 100 corporate clients with cumulative token usage of over 1 trillion。

According to mashable reports, Google Labs launched a new AI tool called CC, designed to replace user morning information browsing habits. The tool has access to Gmail, Google Calendar and Google Drive data, and regularly sends a personalized e-mail bulletin, “Your Day Ahead”, covering calendars, reminders, bills and news. The CC currently has an alternate list for users over 18 years of age in the United States and Canada, with Google AI Ultra and paid subscribers being given priority. Users can..

According to Bloomberg, on Thursday, Google is moving forward with an internal project called "TorchTPU", which aims to enhance the compatibility of its AI software framework, PyTorch, which is the most widely used globally by its artificial intelligence chip, and which refers directly to the ecological moat on which Britain has long relied. According to the source, Google is working closely with Meta, the founder and manager of PyTorch, who wants to enhance bargaining chips with Inverda by lowering the cost of reasoning and diversifying the AI infrastructure. Google's still considering part of..

After the Volcanic Engine FORCE power conference, those involved in the Volcanic Engine responded that “will the Bean Packer model increase the weight of the byte product when it calls App”: “No increase in the weight of the byte product, AI is fully based on the user's instructions.” The person in question explained, for example, demo, a demonstration of the bean bag phone: "In the video of the bean bag assistant, more than the price results have chosen. It's all AI-based, nobody interferes." Besides, Tan, the president of the volcano engine, has been talking about "The Soybean Cellular Assistant has been through a lot of big factories, "the siege."

The TinkStation PGX desktop AI Supercalculator Station has been fully marketed in Kyoto and online. The product provides two versions of 1TB and 4TB, carrying NVIDIA GB10 chip and 128GB memory, with 1 PetaFLOP calculator and only 240 W of power. Support for dual-engineer interconnections, pre-positioning of NVIDIA AI, adaptation of various scenarios such as AI development, scientific research, etc., and provision of three-year warranties based on the Federation ' s preference for the Chinese market。

OpenAI has barely established itself in many U.S. universities, temporarily overcoming university management ' s guard against artificial intelligence, with the ultimate goal of making ChatGPT a common assistant to the next generation of people in the workplace. According to the purchase order, OpenAI sold over 700,000 ChatGPT licences to some 35 public universities for use by students and teaching staff. In contrast, Microsoft, which usually binds its copilot assistants to existing software, has a more modest spread of AI tools in these schools, and teachers tend to use more than students. "The New Wave"

On 18 December, the byte-jumping AI programming tool, TRAE CN, was officially released. The product is used by over 92% engineers inside the byte and has exceeded 6 million registered users in the personal version. Byte in-house practice shows that TRAE contributes more than 431 TP3T to the tremor living service team AI. In response to the four major challenges of performance, deployment suitability, performance tracking, and code security, the TRAE CN enterprise version was fully optimized: supporting the capacity of 100,000 documents, 150 million lines of code; matching the entire scene from the knowledge base to Agent; supporting real-time performance tracking..

Mini Max, a national model company, has been filed by the CVM and heard through the Hong Kong Office and is scheduled to be listed in January 2026. Mini Max was founded in early 2022 and, if successfully listed, will become the fastest AI company in the world from its inception to the IPO, and access to capital markets as “the first share of a globalized AGI”. At present, Mini Max has served 212 million users in over 200 countries and regions around the world, and more than 100,000 business clients have made sustainable profits through fee subscriptions and model calls。

The company LongCat team officially released and opened the SOTA-class Virtual People Video Generation Model, LongCat-Video-Avatar. The model is based on the LongCat-Video base and continues the core design of a model for multitasking, with originals supporting core functions such as Audio-Text-to-Video (AT2V), Audio-Text-Image-to-Video (ATI2V) and video continuation, while fully upgrading the bottom structure to achieve operational integrity, long video stability..

Following Microsoft Azure ' s announcement earlier this month of access to the Kimi k2 model, it is hoped that cooperation between the two sides will go down to the application level. It is known that Microsoft may be working with Kimi this month on the new Agent functionality in response to intense competition from local producers. It is understood that the collaboration will automate Office products using AI's Agent capabilities. This was confirmed to Microsoft and Kimi, who had not responded as of the date of the submission. "The Daily Journal of Science"

Meta Platforms is developing a new image and video-intellectual intelligence model code-named Mango, along with the company ' s next-generation large-language text model. According to sources, the artificial intelligence models were discussed in an in-house employee question and answer session with the company ' s chief product officer, Chris Cox, on 18 December. These models are expected to be released in the first half of 2026。

According to Interface News, CEO Hwang In-hoon of In-hoon of In Weida, on 17 December, revealed that, as part of a plan to invest hundreds of billions of dollars in OpenAI, Yin Weida has not yet paid OpenAI “a penny”. In September of this year, Yen Weida and OpenAI signed an offer of cooperation, and Yen Weida plans to invest $100 billion to OpenAI, with both parties building large data centres. (FIA)

Original link