[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fFsqKOypO9DwL2eIaLzvDiYfxVl-l7G4J3TWDeyRb5Q4":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1570,"年终报道丨“人工智能+”：中国AI开源破局，烟火落地","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">清晨7点，广州李奶奶靠智能手环健康助手查看睡眠数据，顺便用语音询问蚂蚁阿福高血压患者早餐推荐；上午9点，武汉白领小陈佩戴夸克AI眼镜，唤醒AI助手开启骑行导航通勤，一声“夸克拍照”留下所见风景；上午10点，上海张先生用飞书妙记整理会议待办，秒出清单并同步团队；午后2点，深圳金融研究员小周借助千问整理最新财经热点和数据报表，并同步至工作群；傍晚5点半，北京于女士用豆包给儿子生成口算题，错题自动解析讲解；晚上7点，成都王先生通过小爱智能音箱获取家常菜做法；晚上9点，西安视频创作者小宋用剪映AI剪辑工具自动匹配视频BGM、添加字幕；深夜11点，杭州设计师小林用潮汐AI助眠App调节白噪音助眠……\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这组跨越南北的日常场景，正是2025年中国AI产业加速落地的生动缩影。从家庭育儿、职场办公到衣食住行、休闲养生，曾经停留在实验室里的人工智能技术，已褪去高冷外衣，以实用、亲民的姿态渗透到生活之中。这一切变化的背后，离不开中国AI的快速崛起，更折射出中国AI产业从技术突破到生态构建的全面进阶。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">开展“人工智能+”行动是党和国家的一项重要战略部署。“人工智能+”也是对“互联网+”的创新演进。2025年政府工作报告指出，持续推进“人工智能+”行动，将数字技术与制造优势、市场优势更好结合起来，支持大模型广泛应用，大力发展智能网联新能源汽车、人工智能手机和电脑、智能机器人等新一代智能终端以及智能制造装备。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">政策引导与市场活力的结合，让中国AI开源生态进入加速发展期——人工智能开源社区抱抱脸（HuggingFace）数据显示，截至2025年8月，中国开源模型累计下载量已超越美国。全球开发者用脚投票，印证了中国开源技术的全球吸引力。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">当日常烟火与技术革新相遇，中国AI正走出一条兼具商业价值与社会温度的发展之路，同时也在高速发展中面临机遇与挑战的双重拷问。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F60a029e263ce49cea22ddb78535045f6.png\" width=\"null\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-center\">\u003Cspan style=\"color: rgb(187, 187, 187);\">\u003Cspan class=\"ql-cursor\">\u003C\u002Fspan>2025年5月4日，人们在浙江省杭州市文三数字生活街区的AI黑科技市集上体验DeepSeek的人工智能大模型。新华社发（龙巍 摄）\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">技术突破：开源革命引领中国AI产业破局\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">从日常场景到产业深处，国产AI产业能实现如此广泛的落地，核心在于走出了一条不同于美国硅谷的“开源创新”之路。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">不久前，英国《自然》杂志网站发布2025年度十大科学人物榜单，杭州人工智能企业深度求索（DeepSeek）创始人梁文锋入选。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">今年春天，开源大模型DeepSeek-R1的横空出世，这是中国AI技术突破的重要里程碑。此前，行业内外普遍认为，世界上最好的大语言模型掌握在少数几家美国大公司手里。这些大模型闭源，投入高昂、性能优越，由单一机构或商业公司开发维护，有绝对的商业竞争优势，是少数巨头的“专属游戏”。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">DeepSeek打破这一局面。在算力、投入有限的情况下，DeepSeek实现了大模型强化学习技术范式的重大突破，其性能直逼全球最先进的闭源模型。在全球人工智能大模型主流性能测试平台ArtificialAnalysis榜单中，DeepSeek大模型评分达66分，与Gemini、GPT的73分差距大幅缩小。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">技术突破的背后，是中国AI“既开源、又拼速度”的良性竞争局面。如果说技术突破是中国AI的“硬实力”，那么开源模式就是释放价值的“软实力”。DeepSeek以开放之姿，主动打破技术垄断，开放模型架构。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">2025年2月，DeepSeek启动了为期5天的“开源周”活动，通过连续开源多个关键项目，推动AI技术的透明化和生态发展。2025年9月，又于《自然》杂志发表的封面文章中，毫无保留地公开了模型复现的“烹饪配方”，包括关键参数、工程取舍等核心信息，被业内称为“教科书式的透明”。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">《自然》杂志在解释选择梁文锋的理由时，用了一个耐人寻味的词：“可触及的卓越”。坚持开源、免费，意味着让全球研究者和开发者都能站在前人的肩膀上继续创新。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">DeepSeek资深研究员陈德里说，开源是DeepSeek自创立之初就坚定的路线。“在与社区技术开发者交流中，我们收获了很多有意义的反馈，这反过来也推动了DeepSeek自身的发展。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">12月推出的DeepSeek-V3.2系列中，V3.2-Speciale在国际数学奥林匹克、国际大学生程序设计竞赛全球总决赛等四大赛事夺金，进一步缩小了开源与闭源模型的差距。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">“DeepSeek开源之举将使AI像水、电和网络一样触手可及。开源化、轻量化将大力推动‘人工智能+’，广阔场景已经呈现。”浙江大学人工智能研究所所长吴飞说。在ChatbotArena开源大模型榜单中，智谱、月之暗面Kimi、Qwen等中国模型均在前列，彰显了中国开源AI的技术实力。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">美国斯坦福大学《2025年人工智能指数报告》明确指出，中国高性能人工智能模型的数量和质量不断提升，中美人工智能模型性能已近乎持平。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">DeepSeek的开源举措还带来了惊人的生态爆发力：其App上线21天日活跃用户数突破3000万，登顶苹果美国地区应用商店免费App下载排行榜，阿里等多个互联网平台、微信等“国民级”应用、新能源汽车等产业端口纷纷宣布接入，AI芯片厂商、算力产业链企业完成适配，算力需求大幅提升……受到DeepSeek的积极影响，从政府部门、投资机构到产业链上下游，各界信心振奋。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Qwen（通义千问）持续推出全尺寸、全模态大模型系列，累计开源模型超300个，全球下载量突破6亿次，衍生模型数量逾17万个；开源模型MiniMaxM2发布后冲上全球最大模型调用平台OpenRouter趋势榜第一、调用量第三名；火山引擎近日还交出了一份令人瞩目的AI成绩单。截至今年12月，豆包大模型日均token使用量突破50万亿，较去年同期增长超过10倍。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">不久前，《经济学人》报道，硅谷风险投资机构a16z合伙人马丁·卡萨多透露，当前美国AI初创企业中约八成在融资路演时采用的是中国开源模型，而非美国主流产品。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">在全球南方国家看来，开源大模型更是中国帮助世界发展中国家跟上全球技术进步的一份珍贵礼物。这些模型可负担，能够本地化和自由部署。北京大学国际法学院教授斯蒂芬·米纳斯认为，因此，全球南方发展和部署AI的障碍更小、机会更大了。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">赋能于业：AI成为产业提质增效的重要引擎\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">技术的突破最终要以产业变革为落点。2025年，AI已走出实验室，化身渗透创意创作、文化娱乐、软件开发、教育音频等多个领域的重要生产力工具，更迈入“由点及面”的深度应用阶段。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">“当前大模型的应用已进入更为复杂的深水区，从最初的单点应用，逐步深入到各行业的业务主场景之中。”科大讯飞股份有限公司副总裁、研究院院长刘聪的判断，精准勾勒出这一产业演进态势。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">而AI对产业的重塑，主要在于其降本增效、打破壁垒、普惠赋能的核心优势，这一点在创意创作领域尤为凸显。“AI不仅仅是我的另一支笔，它可能是我的伙伴，帮助我走到更多地方，去实现我们目前做不到的创造。”导演陆川近日在第七届海南岛国际电影节作为评委会主席时就坦言，AI不仅仅是一场技术展示，更是一次创作关系的重构实验。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">2024年初，美国OpenAI发布了文生视频模型Sora生成的一个片段，掀起了全球围绕AI视频技术生成的讨论热潮。尽管彼时的AI生成视频效果尚为粗糙，仅能满足用户“尝鲜观看”的需求，但这颗技术种子，却在短短两年内成长为驱动产业变革的参天大树。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">以AI视频技术为突破口，中国产业界开启了从“跟跑”到“并跑”再到“阶段性领跑”的跨越式追赶。2024年下半年，多款中国AI视频生成模型密集亮相，形成群雄逐鹿的竞争态势，推动技术迭代进入“快车道”。MiniMax（稀宇科技）副总裁严奕骏称，初期AI视频行业普遍处于技术探索阶段，生成效果、稳定性都有待提升，但中国企业凭借开源生态的协同优势，快速实现技术突破。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">2024年6月，可灵AI推出DiT视频生成大模型，在视频生成大模型领域实现技术突破，其中可灵2.6实现核心升级，成为业内首批实现AI视频音画同步生成的模型之一。一年半的时间里，可灵AI迭代30余次。快手可灵技术研发负责人温园旭介绍，不久前正式发布的可灵O1依托多模态深度协同能力，就解决了视频生成中角色、场景跨镜头一致性难题，实现架构层面的关键突破。无独有偶，MiniMax也在2025年全力加速产品技术迭代，先后发布了新一代视频生成模型海螺2.3、语音模型Speech2.6等。海螺2.3以强大动态表现力为核心亮点，强化物理表现与指令遵循能力，可呈现更复杂流畅的人物肢体动作，精准度与可控性大幅提升。Speech2.6进一步降低音频生成端到端延迟，复刻目标音色的同时，能生成贴合文本、流利自然的语音。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">随着中国AI视频产业迅速完成从娱乐尝鲜到规模化生产力应用的蜕变，国产模型在细分场景适配、商业化落地能力和性价比上展现出显著优势，不同平台依托自身特色形成差异化竞争格局。陆川导演的实践正是这一趋势的生动注脚，他创办的猿动力已与MiniMax开展深度合作，借助其海螺AI平台开发电影质感级AI漫剧项目，后续还将落地院线电影和精品剧集，以技术与艺术的融合探索创作新可能。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">技术优势也无时无刻不在转化为产业红利。在影视制作领域，传统“高成本、长周期”的痛点被彻底打破：亚马逊PrimeVideo推出的《大卫王朝》中，70多个特效镜头通过AI生成工具制作，大幅缩短特效制作周期、降低资金投入，为影视工业化生产提供全新路径；温园旭表示，尽管AI尚未实现影视全链路替代，但在特效镜头、场景搭建、分镜制作等关键环节，已成为不可或缺的核心辅助工具，推动行业向更高效、更具创意的方向发展。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">温园旭称，短剧行业更是借AI之力实现“弯道超车”，快速生成场景、匹配特效的技术，将单集制作周期从数天压缩至数小时，完美适配行业“快速迭代、海量产出”的核心需求，让中小制作团队也能快速响应市场热点。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AI不仅提升创作效率，更拓宽了文化传播的边界。以往影视剧出海需耗费大量成本进行多语种配音，如今AI可自动生成并匹配不同国家的语言，大幅降低发行成本、提升传播效率。2025年播出的电视剧《国色芳华》，借助AI语音大模型完成多语种配音制作，发行覆盖美国、法国、西班牙等73个国家和地区，成功收获全球观众喜爱。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">在软件开发行业，一场深刻的“生产力革命”正在发生，彻底打破了“懂编程才能做开发”的专业壁垒。用户只需用自然语言描述需求，比如“想要一个包含招聘、绩效、培训功能的人力资源后台系统”，AI就能自动完成代码编写、资料检索、测试运行等复杂工序，20小时左右即可输出成品。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这种低门槛、高效率的开发模式，极大解放了中小微企业、个人工作室的创造力：无论是搭建阶段性任务管理系统、开发简单的手机游戏，还是打造专属软件，都能在几小时到一天内以极低成本实现。尽管产出成果的界面精致度与成熟商业产品存在差距，但其核心功能完全能满足个人或小团队的自用需求，实现了过去难以想象的生产力跃升。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">今年8月，国务院印发《关于深入实施“人工智能+”行动的意见》明确提出，充分发挥我国数据资源丰富、产业体系完备、应用场景广阔等优势，强化前瞻谋划、系统布局、分业施策、开放共享、安全可控，以科技、产业、消费、民生、治理、全球合作等领域为重点，深入实施“人工智能+”行动。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">合肥工业大学北斗导航团队研发出GNSS掩星观测水汽反演（降雨量预测）装置，将北斗导航与人工智能深度融合，成功破解“云中雨”监测难题，相较传统气象雷达、卫星遥感更经济高效，可直接节约千亿元财政投入；在施耐德电气无锡工厂，基于AI技术的热处理数字仿真系统，能够通过算法优化，使单台设备能耗降低25%，氮气消耗减少36%；在铜陵有色金属集团金冠铜业分公司，公司运输物料的皮带运输机自从装上了AI智能巡检系统，能精准识别堵料、异物、跑偏等异常工况，保障了设备的稳定运行，年增经济效益130多万元……\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">技术开始赋能于业，它的深远影响，才刚刚开始。诺贝尔奖虽未设立专门的AI奖项，但近年来经济学奖、物理学奖等诸多学科的突破发展都与人工智能深度相关，AI也早已成为前沿科学研究的重要工具。从小学的人工智能通识教育，到中学的编程实践课程，再到大学的专业人才培养，乃至全社会范围内的AI技能普及，学习和运用人工智能已成为时代趋势。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">服务于人：AI重塑日常，走进千家万户\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这一年，人们遇事先问搜索引擎的习惯正在悄悄改变。当把生活里那些琐碎又麻烦的需求丢向AI助手，它总能拆解成简单明了的答案或指南。从健康守护到教育辅导，从饮食到休闲养生，AI不再是冰冷的工具，而是化身一个个贴近需求的“生活伙伴”，带来触手可及的实际便利，真正浸润到千家万户的烟火中。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">2025年10月，由中国互联网络信息中心发布的《生成式人工智能应用发展报告（2025）》显示，截至今年6月，我国生成式人工智能用户规模达5.15亿人，较2024年12月增长2.66亿人，实现半年即翻番！从“试用”到“常用”再到“天天用”，生成式人工智能正加速融入百姓的日常生活，呈现爆发式增长态势。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">它，可以是你的私人医生——\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">2025年7月，中国科学技术大学附属第一医院医学人工智能联合实验室正式发布“智医随行”大模型。通过本地化部署，深度融合讯飞星火医疗大模型的专病管理路径知识库与医院专科知识库，为医护人员打造覆盖患者“预防—治疗—康复—随访”全周期的AI助手。据了解，目前，“智医随行”大模型已在普外科、骨科、血液内科等六大专科发挥重要助力，覆盖肝切除术后、腰椎间盘突出、膝关节骨关节炎等9类专病，经临床专家审核的专病管理路径，为大量患者提供个性化健康管理服务。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">它，可以是孩子的家庭教师——\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">打开AI辅导软件就能同步当天课本知识点，生成针对性口算练习题，孩子提交后错题立即标注，还附带动画拆解解题思路；孩子遇到难题时，对着AI应用轻声提问，就能快速识别问题并给出清晰讲解，还能通过语音互动引导思考；练习英语口语时，AI可实时纠正发音，模拟真实对话场景，让语言学习更高效有趣。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">它，可以是你的健身搭子——\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">下班回家后，打开AI健身App，输入自己的身高、体重、健身目标和可用器械，系统10秒内就生成了一套30分钟的居家力量训练计划。训练过程中，手机摄像头实时捕捉动作，AI通过姿态识别技术，精准指出“深蹲时膝盖超过脚尖”“平板支撑时臀部过高”等问题，并给出纠正示范。AI健身助手还能联动智能手环记录运动心率、消耗热量，生成周度训练报告，提醒用户合理安排休息与饮食。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">它，不再是一个未知的节点，而是我们正在经历的日常。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">它让我们看见效率的跃升、生活的便利，也让我们思考科技的边界、人文的温度。或许未来，我们将不再纠结于AI能做什么，而是在它的加持下，去探索更辽阔的未知。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">挑战与未来：化“绊脚石”为“磨刀石”\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">从前沿科研的工具革新到全周期教育的体系渗透，AI已全方位融入社会发展的核心脉络，成为驱动进步的关键力量。然而，任何新兴技术的跨越式发展都不会一帆风顺，中国AI产业在收获增长红利、重塑生产生活形态的同时，也不可避免地驶入了机遇与挑战并存的深水区。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">在高速发展的背后，仍面临着技术瓶颈、商业困境、伦理安全等多重挑战，这些问题既是行业发展的“绊脚石”，也是推动产业走向成熟的“磨刀石”，需要全行业冷静审视、共同破解。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">技术与人才层面，创新深水区的攻坚难题凸显。当前大模型虽在多领域取得突破，但在可解释性、安全性、隐私保护等方面仍存在短板，深度学习模型的“黑箱”特性限制了其在医疗、金融等对透明度要求较高领域的深度应用；AI系统面临的数据泄露、模型攻击等风险，对安全技术提出了更高要求；同时，训练超大规模模型仍需消耗巨额计算资源，算力瓶颈尚未完全突破。高端人才的结构性短缺更成为核心制约，顶尖AI人才需兼具深厚的数学、计算机基础与丰富的算法实践经验，培养周期长、总量有限。严奕骏坦言：“AI行业竞争的核心是顶尖人才的竞争，我们团队招聘顶尖大模型算法工程师，往往面试几十位候选人都难以找到完全合适的。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">商业落地层面，盈利难题也亟待破解。市场竞争的白热化导致产品同质化严重，仅AI办公助手赛道就有数十款同类产品，功能趋同迫使企业陷入低价竞争，利润空间被大幅压缩。用户信任度不足也成为阻碍，医疗领域的AI辅助诊断结果、金融领域的智能投顾风险控制能力，仍需长期市场验证，信任壁垒的打破难以一蹴而就。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">伦理安全层面仍面临挑战，AI技术的广泛渗透引发了就业结构调整带来的岗位替代风险、AI决策中的偏见与公平性问题、用户数据隐私泄露风险等一系列问题。在内容创作领域，AI生成内容的版权界定尚不清晰，部分不良信息的传播对社会治理提出新挑战；在自动驾驶领域，事故责任的划分标准仍需完善。尽管全球范围内的AI规范治理已成为共识，欧盟《人工智能法案》与中国《新一代人工智能伦理规范》等政策文件明确了发展的价值导向，但如何将宏观规范落地到具体场景，建立适配不同行业的细分规则，平衡技术创新与风险防控，仍需政府、企业、学界等多方力量共同探索。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">站在2025年的岁末回望，中国AI产业用“开源创新”的突破书写了一份亮眼的答卷，而发展中的挑战更清晰地指明了未来的前进方向：AI的终极价值不在于技术参数，而在于是否能真正赋能于业、服务于人，以普惠姿态融入社会，以责任担当坚守底线。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">展望2026年及更远的未来，笔者认为，中国AI产业的发展或呈现三大趋势。一是“场景深耕”持续深化，AI将从当前的“通用服务”向“垂直细分”进阶。二是“开源生态”持续完善，中国企业将在核心技术研发上持续发力，同时加强与全球开发者的协作，推动开源标准的共建共享，让开源价值惠及更多国家和地区。三是“规范发展”成为共识，政府、企业将共同建立健全AI伦理安全体系，明确数据使用、责任界定等关键规则，为产业发展筑牢安全底线。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">创新工场董事长、零一万物创始人兼首席执行官李开复曾预判，人工智能发展的未来可能类似于苹果和谷歌在智能手机操作系统领域的竞争——美国公司正在构建一个类似苹果iOS系统的封闭生态系统，高价出售；中国的人工智能发展路径则更接近开放且可定制的谷歌安卓操作系统，最终支撑全球大多数的智能手机运行。中国AI企业正在践行的，也是如此——以技术开放为支点，撬动更广阔的市场纵深与更丰富的应用场景，让AI“可触及”。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">从北京于女士的育儿日常，到上海张先生的职场高效，从成都王先生的厨房烟火，到杭州小林的安睡夜晚，AI的价值最终体现在对人的关怀与赋能中。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AI与我们未来会碰撞出怎样的火花？或许现在尚难得出定论。但可以肯定的是，当技术创新与人文关怀深度融合，中国AI必将在全球舞台上书写更精彩的篇章，为人类社会的进步注入更强大的智能力量。（记者朱涵 戴威 王梦瑶）\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">【新闻来源】腾讯网 \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fnews.qq.com\u002Frain\u002Fa\u002F20251231A02G3O00\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F01\u002Ff20d0dbdf64a4bf38fd8bd16feac034c\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F01\u002Fthumbs\u002Ff20d0dbdf64a4bf38fd8bd16feac034c\u002FAI领域.jpg",0,1,48,"2026-01-07 19:07",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3Ab41b8e8d-8e55-41dc-a30b-2575a1dd1b6f%3A0.wav?Expires=1767863431&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=CTQB0fsw86lTkDymVzUFRhwCNfc%3D",38765414,"b41b8e8d-8e55-41dc-a30b-2575a1dd1b6f","2026-01-07 19:06","Year-end Report | \"Artificial Intelligence +\": China's AI Open Source Breakthrough, Everyday Life","\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">At 7 a.m., Grandma Li in Guangzhou checks her sleep data through an intelligent wristband health assistant, and also uses voice to ask Ant's Afu about breakfast recommendations for hypertension patients; at 9 a.m., white-collar worker Xiao Chen in Wuhan wears Quark AI glasses, activates the AI assistant to start cycling navigation for commuting, and says \"Quark take a photo\" to capture the scenery seen; at 10 a.m., Zhang in Shanghai uses Feishu Miaojie to organize meeting to-do lists, instantly generates a list and synchronizes it with the team; at 2 p.m., financial researcher Xiao Zhou in Shenzhen uses Qwen to sort out the latest financial hotspots and data reports, and synchronizes them to the work group; at 5:30 p.m., Yu in Beijing uses Doubao to generate arithmetic problems for her son, and automatically analyzes and explains the wrong questions; at 7 p.m., Wang in Chengdu gets home-cooking methods through Xiaoi smart speaker; at 9 p.m., video creator Xiao Song in Xi'an uses the AI editing tool of KineMaster to automatically match video BGM and add subtitles; at 11 p.m., designer Xiao Lin in Hangzhou uses the Tides AI sleep assistance app to adjust white noise for sleep...\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">These daily scenes across the north and south are a vivid reflection of the accelerated implementation of China's AI industry in 2025. From family education, office work to clothing, food, housing, and entertainment, as well as leisure and health, artificial intelligence that once stayed in laboratories has shed its cold image, permeating into life with practical and approachable attitudes. Behind these changes, it is inseparable from the rapid rise of China's AI, and also reflects the comprehensive advancement of China's AI industry from technological breakthroughs to ecological construction.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Carrying out the \"Artificial Intelligence +\" action is an important strategic deployment by the Party and the state. \"Artificial Intelligence +\" is also an innovative evolution of \"Internet +\". The 2025 government work report pointed out that we should continue to promote the \"Artificial Intelligence +\" action, better combine digital technology with manufacturing advantages and market advantages, support the widespread application of large models, vigorously develop intelligent connected new energy vehicles, AI smartphones and computers, intelligent robots, and other new generation intelligent terminals, as well as intelligent manufacturing equipment.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The combination of policy guidance and market vitality has led China's AI open source ecology into an accelerated development phase—data from the AI open source community HuggingFace shows that as of August 2025, the cumulative downloads of Chinese open source models have exceeded those of the United States. Global developers have voted with their feet, proving the global appeal of China's open source technology.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">When daily life meets technological innovation, China's AI is walking a development path that combines commercial value and social warmth, while also facing the dual challenges of opportunities and challenges during rapid development.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2026\u002F03\u002Fhistory\u002F60a029e263ce49cea22ddb78535045f6.png\" width=\"null\" height=\"null\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-center\">\u003Cspan style=\"color: rgb(187, 187, 187);\">\u003Cspan class=\"ql-cursor\">\u003C\u002Fspan>On May 4, 2025, people experienced DeepSeek's artificial intelligence large model at the AI black tech market in the Wensan Digital Life Street in Hangzhou, Zhejiang Province. Xinhua News Agency photo (Long Wei photo)\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Technological Breakthrough: Open Source Revolution Leads China's AI Industry to Breakthrough\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">From daily scenarios to the depths of the industry, the wide range of applications of domestic AI can be attributed to a unique \"open source innovation\" path different from Silicon Valley in the US.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Not long ago, the British magazine Nature website released the 2025 Top Ten Scientific Figures list, and Liang Wenfeng, founder of Hangzhou AI company DeepSeek, was selected.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This spring, the emergence of the open-source large model DeepSeek-R1 marked an important milestone in China's AI technology breakthrough. Previously, the industry generally believed that the best large language models were held by a few major American companies. These large models are closed-source, require high investment, have superior performance, and are developed and maintained by a single institution or commercial company, giving them absolute commercial competitive advantages, making them exclusive games for a few giants.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">DeepSeek broke this situation. With limited computing power and investment, DeepSeek achieved a significant breakthrough in the paradigm of large model reinforcement learning, and its performance rivals the world's most advanced closed-source models. In the global mainstream performance testing platform ArtificialAnalysis rankings, the DeepSeek model scored 66 points, significantly narrowing the gap with Gemini and GPT's 73 points.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Behind the technological breakthrough is a benign competition situation where China's AI is both open-source and fast-paced. If technological breakthroughs are the \"hard power\" of China's AI, then the open source model is the \"soft power\" for releasing value. DeepSeek, with an open attitude, actively breaks the technical monopoly and opens the model architecture.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In February 2025, DeepSeek launched a five-day \"Open Source Week\" event, promoting the transparency and ecological development of AI technology by continuously open-sourcing multiple key projects. In September 2025, it published a cover article in the journal Nature, openly sharing the \"cooking recipe\" for model replication, including core information such as key parameters and engineering trade-offs, which was called \"textbook-level transparency\" by the industry.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">When explaining the reason for choosing Liang Wenfeng, the Nature magazine used a thought-provoking word: \"Accessible Excellence.\" Persisting in open source and free access means allowing global researchers and developers to stand on the shoulders of predecessors to continue innovating.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Chen Deli, a senior researcher at DeepSeek, said that open source has always been a firm line since the company's inception. \"Through exchanges with community technical developers, we have received many meaningful feedbacks, which in turn have driven the development of DeepSeek itself.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In December, the DeepSeek-V3.2 series saw V3.2-Special win gold in four major international competitions, including the International Mathematical Olympiad and the final of the International Collegiate Programming Contest, further narrowing the gap between open-source and closed-source models.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">\"DeepSeek's open source efforts will make AI as accessible as water, electricity, and the internet. Open source and lightweight will greatly promote 'Artificial Intelligence +', and broad scenarios are already emerging,\" said Wu Fei, director of the Institute of Artificial Intelligence at Zhejiang University. In the ChatbotArena open source large model ranking, Chinese models such as Zhipu, Moonshot Kimi, and Qwen are all at the forefront, demonstrating the technical strength of China's open source AI.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The Stanford University 2025 Artificial Intelligence Index Report clearly states that the quantity and quality of high-performance AI models in China are continuously improving, and the performance of AI models between China and the US has become almost comparable.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">DeepSeek's open source initiatives have also brought surprising ecological explosive power: its App reached over 30 million daily active users within 21 days of launch, topped the Apple US App Store free download rankings, and multiple internet platforms, WeChat and other \"national level\" applications, as well as the new energy vehicle industry, have announced integration. AI chip manufacturers and computing power industry chain enterprises have completed compatibility, and computing power demand has significantly increased... Affected positively by DeepSeek, from government departments, investment institutions to upstream and downstream industries, all sectors are confident.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Qwen (Tongyi Qianwen) continues to release full-size, multi-modal large model series, with more than 300 open-source models cumulatively, and global downloads exceeding 6 billion times, with over 170,000 derivative models; after the release of the open-source model MiniMaxM2, it climbed to the top of the trend list on the world's largest model call platform OpenRouter, ranking third in calls; recently, Volcano Engine also delivered a remarkable AI performance. As of this December, the daily token usage of Doubao's large model exceeded 50 trillion, growing more than 10 times compared to the same period last year.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Not long ago, The Economist reported that Martin Casado, a partner at the Silicon Valley venture capital firm a16z, revealed that approximately 80% of AI startups in the U.S. use Chinese open source models rather than U.S. mainstream products in fundraising presentations.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">From the perspective of developing countries in the Global South, open-source large models are a precious gift from China to help developing countries catch up with global technological progress. These models are affordable, capable of localizing and freely deploying. Professor Stephen Minas of Peking University's School of International Law believes that therefore, the barriers to developing and deploying AI in the Global South are smaller and opportunities are greater.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Empowering Industries: AI Becomes an Important Engine for Industry Upgrading and Efficiency Improvement\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The ultimate goal of technological breakthroughs is industrial transformation. In 2025, AI has left the laboratory, becoming an important productive force tool that permeates creative creation, cultural entertainment, software development, education audio, and many other fields, and has entered a deep application stage from point to surface.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">\"Currently, the application of large models has entered a more complex deep water zone, gradually moving from initial single-point applications into the main scenarios of various industries,\" Liu Cong, vice president and dean of the Research Institute of iFlytek Co., Ltd., accurately outlines this industrial evolution trend.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The reshaping of the industry by AI mainly lies in its core advantages of cost reduction, efficiency improvement, breaking down barriers, and inclusive empowerment, which are particularly evident in the field of creative creation. \"AI is not just another pen for me, but may be a partner that helps me go further and achieve creations that we currently cannot do,\" Director Lu Chuan recently stated as the chairman of the jury at the 7th Hainan Island International Film Festival, emphasizing that AI is not just a technological demonstration, but also an experimental reconstruction of creative relationships.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">At the beginning of 2024, OpenAI released a clip generated by the text-to-video model Sora, sparking global discussions around AI video technology. Although the AI-generated video effects at that time were still rough, only satisfying the need for \"tasting\" viewing, this technological seed grew into a towering tree driving industrial transformation within just two years.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Taking AI video technology as a breakthrough, the Chinese industry has started to move from \"following\" to \"running side by side\" and then to a \"staged leading\" leapfrogging. In the second half of 2024, several Chinese AI video generation models were launched in succession, forming a competitive situation of many strong contenders, pushing the technology iteration into a \"fast lane\". Yan Yijun, vice president of MiniMax (Xiyu Technology), said that initially, the AI video industry was generally in the technical exploration stage, and the generation effect and stability needed improvement, but Chinese companies quickly achieved technological breakthroughs by leveraging the collaborative advantages of the open source ecosystem.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In June 2024, Keling AI released the DiT video generation large model, achieving a technological breakthrough in the field of video generation large models, among which Keling 2.6 achieved core upgrades, becoming one of the first models in the industry to achieve AI video audio-visual synchronization. Within one and a half years, Keling AI iterated more than 30 times. Wen Yuanshu, head of Keling AI's technology research and development, introduced that the recently officially released Keling O1 relies on multi-modal deep collaboration capabilities to solve the problem of cross-shot consistency of characters and scenes in video generation, achieving a key breakthrough at the architectural level. Coincidentally, MiniMax also accelerated product technology iteration in 2025, successively releasing the next-generation video generation model HaiLuo 2.3, the speech model Speech 2.6, etc. HaiLuo 2.3, with its powerful dynamic expression as the core highlight, enhances physical performance and instruction following ability, enabling more complex and smooth character movements, with significantly improved accuracy and controllability. Speech 2.6 further reduces end-to-end latency in audio generation, replicates the target voice while generating natural and fluent voice that fits the text.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">As China's AI video industry rapidly transformed from entertainment tasting to large-scale productivity application, domestic models have shown significant advantages in scenario adaptation, commercialization, and cost-effectiveness, forming a differentiated competitive pattern based on the characteristics of different platforms. Director Lu Chuan's practice is a vivid example of this trend. His company Yuan Dongli has carried out in-depth cooperation with MiniMax, using its HaiLuo AI platform to develop a movie-quality AI comic project, and will later be implemented in cinema films and high-quality TV series, exploring new possibilities of creation through the integration of technology and art.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Technical advantages are constantly being converted into industrial dividends. In the film and television production field, the traditional \"high cost, long cycle\" pain points have been completely broken: in the Amazon PrimeVideo series \"The David Dynasty\", 70+ special effects shots were made using AI generation tools, significantly shortening the special effects production cycle and reducing capital investment, providing a new path for film and television industrialization; Wen Yuanshu said that although AI has not yet fully replaced the entire film and television production chain, it has become an indispensable core auxiliary tool in key links such as special effects shots, scene setup, and storyboarding, driving the industry towards a more efficient and creative direction.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Wen Yuanshu said that the short video industry has also achieved \"overtaking\" with the help of AI, quickly generating scenes and matching special effects, compressing the production cycle of a single episode from several days to several hours, perfectly adapting to the industry's core needs of \"rapid iteration and massive output\", allowing small and medium-sized production teams to quickly respond to market hotspots.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">AI not only improves the efficiency of creation but also expands the boundaries of cultural communication. In the past, the overseas distribution of films and television required significant costs for multilingual dubbing, but now AI can automatically generate and match different languages, greatly reducing distribution costs and improving dissemination efficiency. In the 2025 TV series \"National Beauty and Fragrance\", the AI voice large model was used to complete multilingual dubbing, covering 73 countries and regions including the United States, France, and Spain, successfully winning the affection of global audiences.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In the software development industry, a profound \"productivity revolution\" is taking place, completely breaking the professional barrier of \"only those who understand programming can do development\". Users only need to describe their needs in natural language, such as \"want a human resources back-end system that includes recruitment, performance, and training functions\", and AI can automatically complete complex processes such as code writing, data retrieval, and testing, and output the finished product within about 20 hours.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This low-barrier, high-efficiency development model greatly liberates the creativity of small and micro enterprises and individual studios: whether it is setting up a phased task management system, developing simple mobile games, or creating custom software, they can be realized at a very low cost within a few hours to a day. Although the interface finish of the output results is not as refined as mature commercial products, its core functions can fully meet the self-use needs of individuals or small teams, achieving an unprecedented productivity leap.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In August 2025, the State Council issued the \"Opinions on Deeply Implementing the 'Artificial Intelligence +' Action\", explicitly stating that we should give full play to our advantages in abundant data resources, complete industrial systems, and broad application scenarios, strengthen forward-looking planning, systematic layout, classified measures, open sharing, and safe control, and focus on areas such as science and technology, industry, consumption, people's livelihood, governance, and global cooperation to deeply implement the \"Artificial Intelligence +\" action.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The Beidou Navigation team at Hefei University of Technology developed a GNSS occultation observation water vapor inversion (rainfall prediction) device, integrating Beidou navigation with artificial intelligence, successfully solving the \"cloud rain\" monitoring problem, which is more economical and efficient compared to traditional meteorological radar and satellite remote sensing, directly saving billions of fiscal investments; in the Wuxi factory of Schneider Electric, the AI-based heat treatment digital simulation system can reduce the energy consumption of a single device by 25% and nitrogen consumption by 36% through algorithm optimization; in the Jinguanyue Copper Industry Branch of Tongling Nonferrous Metals Group, since the belt conveyor used for transporting materials was equipped with an AI intelligent inspection system, it can accurately identify abnormal conditions such as clogging, foreign objects, and deviation, ensuring stable operation of the equipment, adding economic benefits of more than 1.3 million yuan annually... The impact of technology empowering the industry has just begun. Although the Nobel Prize has not established a specific AI award, recent breakthroughs in awards such as the Economics Prize and Physics Prize have been closely related to artificial intelligence, and AI has already become an important tool for cutting-edge scientific research. From primary school AI general education, to secondary school programming practice courses, to university specialized talent cultivation, and even AI skill popularization throughout society, learning and applying artificial intelligence has become a trend of the times.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Serving People: AI Reimagines Daily Life, Entering Every Household\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This year, people's habit of asking search engines first is quietly changing. When you throw your琐碎 and troublesome needs into the AI assistant, it can always break them down into simple answers or guides. From health protection to educational tutoring, from diet to leisure and health, AI is no longer a cold tool, but instead becomes a \"lifestyle companion\" that closely meets the needs, bringing tangible convenience and truly permeating into the everyday life of every household.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In October 2025, the China Internet Network Information Center released the \"Report on the Development of Generative AI Applications (2025)\", which showed that by June of this year, the number of generative AI users in China reached 515 million, an increase of 266 million from December 2024, doubling in half a year! From \"trial use\" to \"regular use\" and then to \"daily use\", generative AI is accelerating into people's daily lives, showing a surge in growth.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">It can be your personal doctor—\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In July 2025, the United Laboratory of Medical AI at the First Affiliated Hospital of the University of Science and Technology of China officially launched the \"Zhiyi Suixing\" large model. Through local deployment, it deeply integrates the specialized disease management path knowledge base of the Xunfei Starfire medical large model with the hospital's specialized knowledge base, creating an AI assistant for medical staff covering the entire patient journey from \"prevention - treatment - rehabilitation - follow-up.\" According to the information, the \"Zhiyi Suixing\" large model has already played an important role in six specialties, including general surgery, orthopedics, and hematology, covering nine types of specialized diseases such as liver resection postoperative, lumbar disc herniation, and knee osteoarthritis. The specialized disease management paths reviewed by clinical experts provide personalized health management services for a large number of patients.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">It can be your child's tutor—\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Open AI tutoring software and synchronize the day's textbook knowledge, generate targeted oral arithmetic exercises, and immediately mark wrong questions after submission, accompanied by animated explanations of problem-solving thinking; when children encounter difficult problems, they can ask the AI application softly, and it can quickly identify the problem and give clear explanations, and guide thinking through voice interaction; when practicing English speaking, AI can correct pronunciation in real-time, simulate realistic dialogue scenarios, making language learning more efficient and fun.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">It can be your fitness buddy—\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">After coming home from work, open the AI fitness app, enter your height, weight, fitness goals, and available equipment, and the system generates a 30-minute home strength training plan within 10 seconds. During the training process, the phone camera captures the movement in real time, and the AI, through posture recognition technology, accurately points out issues such as \"knees exceeding toes during squats\" and \"hips too high during planks\" and provides correction demonstrations. The AI fitness assistant can also connect with smart bands to record exercise heart rate, calorie consumption, generate weekly training reports, and remind users to reasonably arrange rest and diet.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">It is no longer an unknown node, but part of our daily experience.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">It allows us to see the leap in efficiency and the convenience of life, and also makes us think about the boundaries of technology and the warmth of humanity. Perhaps in the future, we will no longer question what AI can do, but explore a broader unknown with its support.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Challenges and Future: Turning \"Stumbling Blocks\" into \"Sharpening Stones\"\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">From tool innovations in frontier research to the penetration of the entire education system, AI has been fully integrated into the core of social development, becoming a key force driving progress. However, the跨越式 development of any emerging technology is not smooth sailing. While China's AI industry is reaping the benefits of growth and reshaping the production and lifestyle, it has inevitably entered the deep waters of opportunity and challenge.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Behind the rapid development, there are still multiple challenges such as technical bottlenecks, business difficulties, and ethical security. These issues are both \"stumbling blocks\" for industry development and \"sharpening stones\" for the industry's maturity, requiring the entire industry to calmly examine and jointly solve them.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In terms of technology and talent, the challenges of the innovation deep water zone are prominent. Although current large models have achieved breakthroughs in multiple fields, there are still shortcomings in aspects such as interpretability, safety, and privacy protection. The \"black box\" nature of deep learning models limits their deep application in fields requiring high transparency, such as medicine and finance; AI systems face risks such as data leaks and model attacks, which raise higher requirements for security technology; meanwhile, training ultra-large-scale models still requires a huge amount of computing resources, and the computing power bottleneck has not been completely overcome. The structural shortage of high-end talents has also become a core constraint, with top AI talents needing to have a solid foundation in mathematics and computer science and rich algorithmic practical experience, with a long training cycle and limited total volume. Yan Yijun frankly said: \"The core of competition in the AI industry is the competition for top talents. Our team often interviews dozens of candidates before finding a suitable one for the position of top large model algorithm engineer.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In terms of commercial implementation, the profit dilemma also needs to be solved urgently. The intense market competition leads to serious product homogenization, with dozens of similar products in the AI office assistant sector alone, forcing companies into price competition and greatly compressing profit margins. Insufficient user trust has also become a hindrance, with the accuracy of AI-assisted diagnosis in the medical field and the risk control capabilities of intelligent investment advisors in the financial field still needing long-term market verification, and the breaking of the trust barrier cannot be achieved overnight.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Ethical and security issues still face challenges. The widespread penetration of AI technology has triggered job replacement risks due to restructuring, bias and fairness issues in AI decision-making, and user data privacy leakage risks. In the content creation field, the copyright definition of AI-generated content is still unclear, and the spread of some undesirable information poses new challenges to social governance; in the field of autonomous driving, the standard for accident liability division still needs to be improved. Although global AI regulation and governance have become a consensus, policies such as the EU's AI Act and China's \"New Generation AI Ethics Guidelines\" have clarified the value orientation of development, but how to implement macro-level regulations into specific scenarios, establish detailed rules adapted to different industries, balance technological innovation with risk control, still requires the joint efforts of the government, enterprises, academia, and other parties.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Looking back at the end of 2025, the Chinese AI industry has written an impressive answer sheet with the breakthrough of \"open source innovation,\" and the challenges in development have more clearly pointed out the direction of future progress: the ultimate value of AI does not lie in technical parameters, but in whether it can truly empower industries, serve people, integrate into society with a popular attitude, and uphold responsibility and safeguard the bottom line.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Looking ahead to 2026 and beyond, I believe that the development of China's AI industry may show three trends. First, \"scenario deepening\" will continue to deepen, and AI will move from the current \"general service\" to \"vertical specialization.\" Second, the \"open source ecosystem\" will continue to improve, and Chinese companies will continue to invest in core technology R&D, while strengthening collaboration with global developers to promote the co-construction and sharing of open source standards, so that the value of open source can benefit more countries and regions. Third, \"regulated development\" will become a consensus, the government and enterprises will jointly establish and improve the AI ethics and safety system, clarify key rules such as data usage and liability definition, and build a safety bottom line for industrial development.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Li Kaifu, chairman of Innovation Works and founder and CEO of Zero One Ventures, once predicted that the future of AI development may be similar to the competition between Apple and Google in the smartphone operating system field—U.S. companies are building a closed ecosystem similar to Apple's iOS system, selling it at high prices; China's AI development path is more like the open and customizable Google Android system, ultimately supporting the operation of most smartphones worldwide. The practices of Chinese AI companies are exactly like this—using technical openness as a lever to drive a broader market depth and richer application scenarios, making AI \"accessible.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">From Ms. Yu's parenting routine in Beijing, to Mr. Zhang's efficient workplace in Shanghai, from Mr. Wang's kitchen scenes in Chengdu, to Mr. Xiao Lin's peaceful night in Hangzhou, the value of AI is finally reflected in the care and empowerment of people.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">What sparks will AI create with us in the future? It may be difficult to reach a conclusion now. But one thing is certain—when technological innovation and humanistic care are deeply integrated, China's AI will certainly write a more splendid chapter on the global stage, injecting a stronger intellectual power into the progress of human society. (Reporter Zhu Han, Dai Wei, Wang Mengyao)\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(136, 136, 136);\">[News Source] Tencent News \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fnews.qq.com\u002Frain\u002Fa\u002F20251231A02G3O00\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(136, 136, 136);\"> https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1RTDQy?ocid=BingHp01&cvid=6936317f054647a2afcd53fafcde084a&ei\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(136, 136, 136);\">（This article is forwarded by this site to provide readers with more information, and the content does not constitute investment or consumption advice. If there are any doubts about the facts of the article, please verify with relevant parties. The views of the article are not the views of this site, and are for reference only.）\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A7dfee0f2-e040-4e73-b5a3-769aca66968c%3A0.wav?Expires=1774838432&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=utJtbmS3P3rf6dNuLmAWb6yJM8o%3D","7dfee0f2-e040-4e73-b5a3-769aca66968c",17510336]