OpenAI将在几个月内发布GPT-5,而不是单独的O3型号
Openai首席执行官Sam Altman在2月13日在社交媒体上说,Openai已更新了其针对GPT-4.5和GPT-5的产品路线图。 OpenAI将于下一步发布GPT-4.5,其中包括产品代码名称Orion,这将是OpenAI的最后一个无思想的链型。接下来,OpenAI将在ChatGpt和API服务中携带新的GPT-5。 GPT-5将集成公司的多种技术,包括O3模型技术。最初计划单独推出的O3模型将不再单独发布。奥特曼说,GPT-4.5将在几周内发布,GPT-5将在几个月内发布。
评论:DeepSeek于1月底发布了R1模型,并声称该模型的功能是针对OpenAI O1模型的基准测试的,OpenAI被认为承受了很大的压力。 R1启动后,OpenAI按计划于1月31日本地时间发布了下一代O1的迷你版。但是O3米尼的发布似乎并不能减轻Openai对市场竞争的焦虑。在该模型发布当天,Openai在互联网平台上进行了一轮问答。山姆·奥特曼(Sam Altman)在问答中承认,Openai的领先职位不如前几年。
Tsai Chongxin确认阿里巴巴和苹果合作
2月13日,在阿联酋迪拜举行的世界各国政府峰会2025峰会,阿里巴巴联合创始人兼董事会主席,泰·钟X(Tsai Chongxin)回应了阿里巴巴与苹果AI之间合作的谣言。他说,苹果需要与中国AI公司合作,以开发中国iPhone的AI功能。苹果一直很挑剔,他们与中国的几家公司进行了交谈。最后,他们选择与阿里巴巴开展业务。
评论:对于苹果而言,这一举动可能是它扭转中国市场销售下降的重要一步。对于阿里巴巴来说,这是其“ AI驱动”策略的重要突破,将有助于促进其技术转型。这种合作不仅填补了苹果生成的AI策略的关键地图,而且还为阿里巴巴提供了AI策略方向的新想象力。新闻发布后,2月13日下午,阿里巴巴的香港股市上升至9%,其股价高达三年。此外,记者还了解到,苹果仍在与百度合作开发iPhone的AI技术,这意味着苹果已经选择了多个合作伙伴。
百度将开放下一代文学和心脏模型
2月13日,百度宣布,从4月1日起,所有Wenxin Yiyan都将是免费的,并且将启动深入的搜索功能。 2月14日,百度宣布将在接下来的几个月内启动Wenxin Mockup 4.5系列,并将从6月30日正式开放。
评论:百度以前是大规模封闭源的坚定支持者。去年,百度主席罗宾·李(Robin Li)表示,大型模型的开源不是很有意义,封闭的源模型将继续领导能力。只有带有业务模型的封闭源模型才能收集人力和财务资源。中国和美国最强大的基本模型已关闭。来源。但是,情况现在已经改变。 DeepSeek打开了基于性能的Model R1,该模型R1基准了OpenAI O1,并在前几个大型模型中具有开源模型。如果某些无法产生更好性能的大型模型继续关闭,制造商继续依靠API(接口)来收取费用以获得收入,则该模型将很难继续。
马斯克想以974亿美元的价格收购Openai
2月11日,由马斯克(Musk)领导的一个投资联盟提议收购一个以974亿美元(约合7119亿元人民币)控制OpenAI的非营利组织。根据收购计划,在交易结束后,马斯克将与其子公司转移OpenAI。 xai合并。作为回应,OpenAI首席执行官Sam Altman从X平台上的远处做出了回应,“不,谢谢您,但是如果您愿意,我们愿意以97.4亿美元的价格收购Twitter。”
评论:此前,在2024年10月,Openai完成了66亿美元的融资,估值为1570亿美元。同时,Openai承诺在2026年底之前完成营利性重组,将OpenAI现有的营利性组织转变为特拉华州的公共利益企业(PBC),并继续维护非营利组织和A供 - 利润组织。状态。这次,马斯克(Musk)对OpenAI的某些非营利资产的不请求竞标极大地提高了Altman转换为“非营利性 +营利性”计划的复杂性。由于Altman的重组计划,最困难的问题之一是如何重视非营利性部分。阿尔特曼在巴黎AI峰会上说,马斯克试图通过积极竞标Openai来推迟公司的开发过程。 “
Openai首席执行官说,AI每年的成本下降约10次
2月9日,OpenAI首席执行官Sam Altman发表了一个长期的博客,讲述了他关于人工智能经济学的三个观察:AI模型的情报水平与培训和运营所消耗的资源大致相同。对数是成比例的;使用特定水平的AI的成本每12个月下降约10次,而价格较低将导致更多使用;智能线性增长带来的社会经济价值本质上是超级指数的,因此结果是,对AI的投资将在未来继续表现出指数增长。
评论:Sam Altman多次强调AI成本正在下降。从使用OpenAi产品的成本来看,每个令牌的价格从2023年初的GPT-4到2024年中期的GPT-4O下降了约150倍。预计成本下降将推动应用程序的爆发。
苹果机器人早在2028年就已经大量生产,目前处于POC早期阶段
On February 13, Tianfeng International Securities analyst Ming-Chi Kuo posted on X that Apple is exploring humanoid and non-humanoid robot projects for its future smart home ecosystem, and these products are currently in the early proof of concept (POC) stage. There is still a distance from POC to formal product development. According to the current progress and conventional product development cycle, Apple robots may enter the mass production stage until 2028 or later. He also pointed out that Apple disclosed some of its research results in the early stages of POC, contrary to the normal state of attaching importance to confidentiality, which may be to recruit talents.
Comment: Apple Machine Learning Research Center previously released a video and a 12-page paper introducing its new robot prototype. This desktop robot looks similar to Pixar's classic character Luxo Jr. Recently, Apple has also posted a series of positions related to home robot research and development on its recruitment page, mainly for professional talents with backgrounds in mechanical engineering, electronic engineering, artificial intelligence and机器学习。 From the current public information, it can be seen that the Apple robot team is led by Vice President of Technology Kevin Lynch.
EU will invest 50 billion euros to promote artificial intelligence research and development
On February 12, the EU announced the launch of the "Invest AI Plan" at the AI Action Summit in Paris, which will invest 50 billion euros of public funds in artificial intelligence research and development in the future, and jointly form a investment计划总比例为2000亿欧元。 The plan includes 20 billion euros to build four AI Gigafactories, which will be equipped with about 100,000 latest generation AI chips, four times the amount of AI factories currently under construction.这些超级工厂将为欧洲初创企业和科学家提供强大的计算能力,以推动复杂的AI模型的开发。
Comment: European Commission President von der Leyen stressed that Europe does not lag behind China and the United States in cutting-edge technology fields, and will build competitiveness by focusing on advantageous areas. The plan will focus on supporting the digital transformation of small and medium-sized enterprises, while promoting the application of AI technology in key industries such as medical care and energy.
机构预测数据中心资本支出将继续增长
On February 11, market research firm TechInsights released a report saying that in 2025 and beyond, the data center semiconductor industry will see several trends. First, data center capital expenditure will continue to grow, and top super-large cloud service providers will invest about US$320 billion in AI in 2025. Second, the cost of inference for large language models will decrease.第三,世界各地的大多数数据中心基础架构都超过了六年,旧服务器将迎来大规模更新。 In addition, in 2025, software and hardware collaborative design is crucial to the increasingly large and complex AI models.
Comment: The low cost of DeepSeek training V3 has not allowed overseas technology giants to reduce capital expenditures, and cloud manufacturers are still investing in building data centers. The research institute also believes that DeepSeek-R1 is actually a way to make the prospect of artificial intelligence move from training to reasoning. In the future, competition will no longer be who can build better models, but how to reduce the cost of each query as much as possible.因此,推理成本将下降。此外,一些行业内部人士认为,向人工智能的前景转变将增加对推理的需求,这将带来对计算电源芯片的强烈需求,这可能是技术巨头仍在部署数据中心的原因。
大脑计算机界面启动阶梯医学完成了3.5亿元B圆金融资
2月12日,上海Jietie Medical宣布已完成3.5亿元S系列融资,这是中国可植入的脑部计算机界面行业历史上最大的融资。据了解,这一轮融资是由Qiming Venture Capital,Aobo Capital和Eli Lilly Asia Fund和其他机构领导的。天使巡回投资机构Yuanlai Capital继续进行跟进,Qifeng Capital担任财务顾问。 The financing funds will be mainly used to accelerate product clinical trials, new generation technology research and development, and the construction of medical-grade MEMS production bases, providing strong support for the commercialization of brain-computer interface technology.
Comment: According to the path of brain signal acquisition, brain-computer interface technology is divided into non-invasive (extracerebral), invasive (intracerebral cortex) and semi-invasive. At present, most Chinese brain-computer companies choose the non-invasive acquisition technology route, and invasive brain-computer interface technology is still in its early stages of development. Domestic invasive brain-computer interface company NeuraMatrix officially announced the completion of a Series A financing of hundreds of millions of yuan in 2022, and Boruikang Medical Technology completed a Series D financing in November 2024.
Tencent Cloud将建立其第一个中东数据中心
On February 9th local time, at the Saudi Arabian technology event LEAP2025, Tencent Cloud announced that it will build the first Middle East data center in Saudi Arabia. At the same time, in conjunction with the needs of local industrial innovation, it will invest more than US$150 million in the Middle East in the next few years.用于基础设施,资源和相关的投资建设。数据中心预计将于2025年推出,并将采用双重可用性区域模式,以提供高质量的云服务,例如弹性计算,存储,安全性,AI等。
评论:中东的一些国家正在发展石油以外的其他行业,数字经济是主要的一步,其中人工智能受到了高度重视。此前,Tencent Cloud与阿布扎比文化和旅游局合作,使用数字智人技术来创建数字虚拟图像“ Karifa”。中东的当地机构正在增加他们对人工智能部署的投资,这也意味着对计算能力的需求。 2023年,阿联酋阿布扎比理工学院创新研究所TII推出了大型Falcon 180B,那一年,阿联酋G42 G42推出了阿拉伯人人工智能语言模型JAIS。随着国内云计算市场的增长率下降,云制造商选择出国。
克里斯蒂(Christie
2月10日,外国媒体报道说,最近有3,000多位艺术家写信,抗议英国拍卖行克里斯蒂(Christie)的人工智能艺术拍卖。该报告说,佳士得的第一次人工智能艺术拍卖称为“增强的情报”,该拍卖是大型拍卖行进行的首次人工智能拍卖。拍卖将由艺术家使用人工智能模型进行。 Holding in New York.
克里斯蒂(Christie)的数字艺术总监妮可(Nicole)销售吉尔斯(Nicole Sales Giles)表示,人工智能“不能替代人类的创造力……它增强了人类的创造力。”然而,艺术家在给克里斯蒂的一封信中说,许多拍卖的艺术品都是使用人工智能模型创建的,据称未经其许可就受到了受版权保护的作品的培训。
评论:拍卖强调了创意行业对AI的复杂辩论,艺术家在使用AI以及该技术是否将成为创作过程中的关键工具方面有所不同。 Some voices in the art world also support the auction, and some even used the text and patterns on the protest letter to create digital art.
AI儿科医生“担任他们的职位”,并与13位儿科专家完成了有关困难案件的咨询
2月13日,北京儿童医院隶属于国家儿童医学中心的首都医科大学,在中国举行了“ AI儿科医生 +多学科专家”的第一次双学科平行的多学科咨询。 13来自耳鼻喉科,头颈外科,肿瘤外科,肿瘤学内科,神经外科手术等的专家进行了对患有颅底肿瘤的儿童伴有伴有抽动症状的儿童的多学科咨询。 AI儿科医生也在工作。工程师将在患者的主要投诉和病历中输入模型后,AI儿科医生提出的建议与专家小组的咨询结果非常一致。 "AI Pediatrician" was jointly developed by Beijing Children's Hospital, Baichuan Intelligent and Pediatric Health Technology. It was launched on January 18 and officially "taken on the job" on February 13.
评论:AI医生整合了来自北京儿童医院的300多名著名儿科专家的临床经验,并从专家那里获得了数十年的高质量医疗记录。通过结构化的临床推理范式培训,他们建立了三维知识,涵盖了儿童常见和困难的疾病。系统。 In clinical scenarios, "AI pediatricians" can help doctors obtain the latest scientific research results or authoritative guidelines for complex clinical problems such as interdisciplinary and cross-disciplinary diseases, and can also assist doctors in the diagnosis and treatment of difficult diseases. AI medical care is important for general artificial intelligence. One of the application areas, the employment of AI doctors is also expected to promote the popularization and universal access of high-quality medical services.