News

11月12

  • Human intelligence locked up in 2021, OpenAI CEO predicts AGI can be realized in 5 years | 10,000-word fact sheet

    09:39

    OpenAI CEO Sam Altman predicts that AGI could be here within five years, noting that human intelligence may be stagnating. the spread of AI has changed the way information is accessed and problems are solved, while creating the problem of AI spam. altman emphasizes the importance of open-source and no-code tools and believes that inference models will drive progress in scientific research. ai startups will need to face rapid technological advances to explore long-lasting market opportunities. advances and explore long-lasting market opportunities. (Altman)

    Share to
  • Another Domestic GPU Unicorn to Launch IPO, Valued at Over $25.5 Billion 

    09:38

    Moore Threads, a domestic GPU startup founded in 2020, has undergone a shareholding reform and is planning an IPO on Kechuan Board.The company focuses on the development of full-featured GPU chips and solutions to support AI computing and digital transformation of industries. As of April 2024, the valuation reached 25.5 billion yuan, and it was selected in the 2024-Hurun Global Unicorn List. (Direct IPO)

    Share to
  • A piece of artwork sells for nearly 8 million dollars! No need for factory labor, robots find new way to make a living 

    09:38

    Portrait of Turing created by humanoid robot Ai-Da sold for $1.1 million, auction proceeds to fund further R&D Ai-Da, the first hyper-realistic robotic artist with hand-eye synergy and other capabilities, has sparked a discussion about the ethics of AI and the boundaries of creativity by auctioning off its paintings. (Wisdom)

    Share to
  • Big models can't even multiply? A new series of studies challenges AI reasoning once again

    09:38

    The study found that AI macromodels did not master traditional computational rules when dealing with math problems, but instead used a form of statistical pattern matching. Such models are more like complex pattern matching than logical reasoning. Although COT technology was introduced to improve reasoning ability, it still relies on pattern recognition rather than true mathematical understanding.AI models still have to make breakthroughs in processing computational rules and reasoning ability. (AI Future Pointing North)

    Share to
  • o1/Claude Collective Flop! Tao Zhexuan and 60+ other top mathematicians join forces to propose new math benchmarks

    09:38

    Epoch AI has launched FrontierMath, a new math benchmark that challenges the mathematical reasoning of top big models. Models including GPT-4o and others have solved less than 2% problems on the benchmark.Designed by more than 60 leading mathematicians, the set of benchmarks covers areas ranging from number theory to algebraic geometry and is extremely difficult. In the future, the study will add questions and optimize the evaluation process. (Quantum Bits)

    Share to
  • Taiwan official: TSMC can't produce 2nm chips overseas at this time

    09:38

    Taiwan's economic department head Guo Zhihui said, TSMC 2-nanometer process is currently subject to regulatory restrictions can not be produced overseas, the future will be produced in the United States, but the most advanced technology will stay in Taiwan. He believes that without TSMC, the United States is difficult to obtain competitive chips, and said that the United States to set up factories will bring a large number of jobs, 2 nm is expected to be used for U.S. production after 2030 is still to be assessed. (Observer)

    Share to
  •  OpenAI Head of Product: Today's Models Are Limited by Evaluation Methods

    09:37

    Kevin Weil, Chief Product Officer at OpenAI, and Mike Krieger, Chief Product Officer at Anthropic, discuss a number of topics central to the field of artificial intelligence. They discussed the limitations of current AI models, emphasizing that these limitations lie more in the evaluation methodology than in the level of intelligence itself, noting that the role of the product manager is undergoing a shift from the traditional role to that of a research-based product manager, which requires them to acquire new skills such as writing evaluation criteria and fine-tuning models. The two product leaders also looked to the future of AI, suggesting that "proactive...

    Share to
  • Former Head of AI Platform at ByteDance Becomes Partner at Sanjee Technologies

    09:36

    According to Z Potentials, Pan Xin, a former Google brain first research software engineer, former head of ByteDance's AI platform, and co-founder of the head AI modeling company, has been appointed as a partner of Flash Pole Technology, in charge of the development of AI-related technologies. According to the report, he will set up and lead a professional and efficient R&D team to create a killer AI product that recommends information and services based on multimodal AI interactions, real-time scene analysis, and other technologies based on users' real-life scenarios and habits. According to the report, Pan Xin joining Flash Pole Technology is expected to promote...

    Share to
  • Conan AI Raises Multi-Million Dollar Angel+ Series Funding from Tahoe Capital

    09:36

    November 11 news, Conan (Guangdong Hengqin) Intelligent Technology Co., Ltd. completed a multi-million dollar angel + round of financing, the investor is Taihe Capital, Huafeng Capital as a long-term strategic financial adviser, this round of financing funds will be used for product manufacturing, supply chain optimization, marketing and to strengthen the construction of the algorithms and arithmetic power of AI products.

    Share to
  • AI healthcare and brain science company Virtual Reality Technologies receives tens of millions of dollars in Series A investment

    09:36

    November 11 news, AI medical and brain science company Hangzhou Virtual Reality Technology Co. completed the A round of financing. This round of financing was led by Xingtuo Capital and followed by AIalpha Ventures and other participants, with a financing scale of tens of millions of yuan. The financing will be mainly used to accelerate the domestic and overseas medical product development, C-end and expert market team construction. Next, the company will further consolidate its leading position in the field of digital treatment of brain diseases in China, explore more innovative business models, and provide users with more excellent products and services.

    Share to
  • After acquisition by SoftBank, AI chip company Graphcore expands by 20% employees

    09:36

    Graphcore, the chip design company acquired by SoftBank Group in July, has added 75 new jobs globally and increased its workforce by 20%, underscoring its parent company's ambitions in the semiconductor industry. The new round of hiring will bring Graphcore's workforce to about 450 people.Graphcore said in a statement that the new positions, which include wafer design and data center engineering roles, will be based in the U.K., Poland and Taiwan, China. (Tibco)

    Share to
  • The number of AI organizations in China is nearly 1.9 million

    09:34

    November 12 news, the General Administration of Market Supervision of the national organization unified social credit code data service center announced today, as of October 31 this year, the national "artificial intelligence" organization has reached nearly 1.9 million households. The latest data show: artificial intelligence-related organizations in the southeast coastal region is particularly concentrated, accounting for up to 57%, of which the Guangdong region to nearly 280,000 households led the country, occupying the national total of 14.7%. Hainan Province in recent years with an average annual growth rate of 55%, the total number of more than 90,000, ranking among the forefront of the country's large artificial intelligence province. In terms of industry distribution, the two major fields of science and technology, information technology...

    Share to
  • OpenAI's bet on 1X unlocks new scenarios with humanoid robot as chef

    09:34

    In August, OpenAI-bet robotics company 1X Technologies officially unveiled NEO Beta, a prototype humanoid robot designed for the home. the robot recently unlocked a new scenario - the kitchen - as NEO and Netflix chef Nick DiGiovanni engaged in a domestic cooking showdown between NEO and Netflix chef Nick DiGiovanni, competing to make the perfect medium rare steak. 1X's team says the robot's cooking capabilities won't be immediately available to the first NEO users, and they want to make sure that NEO gains experience with safer tasks before handling sharp or hot objects.

    Share to
  • SeedEdit, a universal image editing model for byte beanbags, opens for beta testing, allowing users to easily change the image in a single sentence

    09:34

    November 11 news, bytes in the beanbag big model team official website announced the latest universal image editing model SeedEdit. SeedEdit support a sentence easy to change the map, including retouching, dressing, beautification, transforming the style, add and delete elements in the specified area and other types of editing operations, through a simple natural language can be driven by the model to edit any image. Currently, the model has been opened for testing in the PC terminal of Doubao and the web terminal of Dream. Users can generate images in Doubao, click the continue editing button, enter simple text commands to adjust the background or subject of the image, to realize a sentence to change the picture.

    Share to
  • UK 20% Doctors Are Using ChatGPT for Medical Diagnosis, Medical Advice and More

    09:34

    Doctors.net.uk, the UK's largest professional doctor service platform, has done a survey on the use of ChatGPT (a product of OpenAI), Bing AI (a product of Microsoft) and Google Gemini (a product of Google Inc.) to 1,000 professional doctors. The results of the survey showed that 20% doctors in the UK are using generative AI to assist in their daily work, with ChatGPT having the highest usage rate at 16%, Microsoft's Bing coming in second and Google's Gemini coming in third.

    Share to
  • Google to launch stand-alone 'Google Gemini' application

    09:34

    Google is launching an Apple Google Gemini app. iPhone users can now only access Gemini through a dedicated tab in the Google app. Google is testing a standalone "Google Gemini" app on the Apple App Store that will offer more AI features than the current Google app. Google is testing a standalone "Google Gemini" app on the Apple App Store that will offer more AI features than the current Google app.

    Share to
  • URAvatar: Generate high-fidelity virtual avatars and enhance their visualization through cell phone scanning

    09:33

    URAvatar technology generates high-fidelity virtual avatars and enhances their visualization through cell phone scanning, allowing users to drive and adjust the avatar in real time. The technology employs a learnable radiative transfer model for real-time rendering and light migration, which gives new possibilities for virtual avatars, and allows users to independently control the gaze direction and neck movement of the avatar, which enhances the virtual interaction experience. The paper is available at: https://arxiv.org/html/2410.24223v1 https://junxuan-li.github.io/urgca-websit...

    Share to
  • Kunlun Wanwei's SkyReels AI sketch platform will officially launch in the US on December 10th

    09:33

    SkyReels, an AI sketch platform, will soon be officially launched in the U.S., bringing a new intelligent sketch experience to North American audiences. The platform provides content creators with powerful creation tools through innovative technologies and features, while lowering the threshold of AI skit creation so that non-professional users can easily get started. Official website address: https://skyreels.ai/beta

    Share to
  • Nous Research Launches Nous Chat Chatbot, Based on Meta Llama 3.1 Fine-Tuned "Hermes 3 - 70B" Model

    09:33

    AI startup Nous Research has unveiled its Nous Chat chatbot based on the "Hermes 3 - 70B" model fine-tuned for Meta Llama 3.1, which it claims is deeply improved and human-centered in a number of ways. Hugging Face has opened up the model family for developers to preview. The company also previewed the Nous-Hermes-2-Mixtral-8x7B-...

    Share to
  • Technische Universität Wien: Self-learning robot that cleans washbasins in imitation of humans

    09:33

    Researchers at the Technical University of Vienna have developed a self-learning robot that can mimic a human cleaning a washbasin. Traditional robot programming is complex for modeling washbasins, and the team trained the robot with a combination of human demonstrations and haptic data, a learning method that can be applied to a wide range of industrial production tasks. Learning in a similar way to humans, the team developed special cleaning tools, generated data and processed it through several demonstrations so that the robot understands the correct way to clean, and also used innovative data processing strategies. In the future, the robot could be mounted on a mobile platform, becoming a powerful assistant on the shop floor and able to pass on its experience to other robots.

    Share to

Search