[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-detail:ZX20250129_68047754":3},{"code":4,"data":5},0,{"unicode":6,"title":7,"categoryId":8,"categoryName":9,"industryId":10,"industryName":11,"origin":12,"originAuthor":13,"originRemark":14,"originUrl":15,"pic":16,"content":17,"numRead":18,"numLike":4,"createTime":19,"hadFav":4,"hadLike":4,"status":4,"memberUnicode":20,"companyLogo":21,"companyName":22,"corporateVideoUrl":23,"collected":4,"corporate3dCover":24,"corporate3dUrl":25,"corporate3dId":26,"companyUnicode":27,"focusFlag":28,"exhibitionMainInfoVO":29},"ZX20250129_68047754","国货DeepSeek引发AI风暴的六点思考","1242381412612444162","行业动态","1245513623956107264","其它行业",1,"","http:\u002F\u002Ftech.ifeng.com\u002F","https:\u002F\u002Ftech.ifeng.com\u002Fc\u002F8gVdf0wov2s","https:\u002F\u002Fx0.ifengimg.com\u002Fucms\u002F2025_05\u002F95B8DDDCB738B3918B0ABCB34B7592E3A2D72619_size1081_w1080_h1969.png","\u003Cdiv class=\"index_articleBox_6mBbT\" style=\"height:auto\">\n \u003Cdiv class=\"index_text_D0U1y\">\n  \u003Cp>国货deepseek引发AI风暴的六点思考：\u003C\u002Fp>\n  \u003Cp>\u003Cstrong>1\u003C\u002Fstrong>\u003C\u002Fp>\n  \u003Cp>大语言模型方面，中美在大语言模型领域已进入并跑阶段。OpenAI长期保持的技术壁垒被低成本颠覆，开源和开放的策略彻底改变了行业规则，这是引发广泛关注的主要原因。DeepSeek的出现不仅打破了技术垄断，推动了全球AI技术的普及。\u003C\u002Fp>\n  \u003Cp>\u003Cstrong>2\u003C\u002Fstrong>\u003C\u002Fp>\n  \u003Cp>现在说，全面超越为时尚早。AI创新生态的复杂性，远非单一的大语言模型所能概括。DeepSeek的成功恰恰证明了全球协同创新的重要性。技术封锁在全球化背景下意义有限，逆全球化不符合科技发展的趋势。AI产业的进步依赖于全球合作，而非孤立竞争。\u003C\u002Fp>\n  \u003Cp>\u003Cstrong>3\u003C\u002Fstrong>\u003C\u002Fp>\n  \u003Cp>英伟达应该会继续跌。随着更多AI公司转向更具成本效益的训练方案，前期购买大量卡的公司要消化一段时间，英伟达的市场需求增长将受到挑战。推理方面，国产化推理卡已经能够满足大部分需求，除非算力需求大幅提升，否则英伟达的市值难以维持当前水平。预计其市值可能进一步下跌30%，跌破2万亿美元。\u003C\u002Fp>\n  \u003Cp>\u003Cimg class=\"empty_bg\" src=\"https:\u002F\u002Fx0.ifengimg.com\u002Fucms\u002F2025_05\u002F95B8DDDCB738B3918B0ABCB34B7592E3A2D72619_size1081_w1080_h1969.png\" src=\"data:image\u002Fpng;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABAQMAAAAl21bKAAAAA1BMVEXy8vJkA4prAAAACklEQVQI12NgAAAAAgAB4iG8MwAAAABJRU5ErkJggg==\" style=\" width: 640px; height: 1166px;\">\u003C\u002Fp>\n  \u003Cp>\u003Cstrong>4\u003C\u002Fstrong>\u003C\u002Fp>\n  \u003Cp>LLM本身还是工程创新，不是基础科学突破。其技术壁垒并不高，主要依赖数据和算力的持续迭代。这也是OpenAI不愿完全开源的原因——投入巨大但回报尚未完全实现。deepseek较短时间内又做了一次拉齐，对AI产业是好事。DeepSeek在短时间内实现了技术对齐，推动了AI产业的进步。未来，更深层次的技术突破可能会加速到来。美国的AI生态依然强大，谷歌的稳定表现和持续积累值得关注，不应只聚焦于OpenAI。\u003C\u002Fp>\n  \u003Cp>\u003Cimg class=\"empty_bg\" src=\"https:\u002F\u002Fx0.ifengimg.com\u002Fucms\u002F2025_05\u002FE295F6517DFD8CB335EB06BAA3D4395A589D3578_size219_w1080_h480.png\" src=\"data:image\u002Fpng;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABAQMAAAAl21bKAAAAA1BMVEXy8vJkA4prAAAACklEQVQI12NgAAAAAgAB4iG8MwAAAABJRU5ErkJggg==\" style=\" width: 640px; height: 284px;\">\u003C\u002Fp>\n  \u003Cp>\u003Cstrong>5\u003C\u002Fstrong>\u003C\u002Fp>\n  \u003Cp>中国如何从并跑到领跑。中国要实现从并跑到领跑的跨越，必须跨过大语言模型，提前和加快布局下一代AI技术。当前的隐忧在于创新机制的不足。虽然企业是技术发明和产品创新的主体，但在前沿理论发现方面仍显薄弱。领跑需要依靠理论构建和科学发现，而中国的研究体系与企业创新体系之间仍存在壁垒。推动顶尖科学家与创新企业家的协同合作，形成创新联动机制，是实现持续突破的关键。\u003C\u002Fp>\n  \u003Cp>\u003Cstrong>6\u003C\u002Fstrong>\u003C\u002Fp>\n  \u003Cp>接下来的AI发展将更加精彩，值得期待。随着技术的不断进步和全球合作的深化，AI将在更多领域展现出巨大的潜力。\u003C\u002Fp>\n \u003C\u002Fdiv>\n \u003Cspan>\u003C\u002Fspan>\n \u003Cdiv class=\"index_end_1O-ki\">\u003C\u002Fdiv>\n\u003C\u002Fdiv>",7,"2026-09-21 10:09:43","2299021137","https:\u002F\u002Fimages6396289.oss-accelerate.aliyuncs.com\u002F2299021137\u002F207k1en0gl\u002F39556b555c5294a73c851a6d064c914.png","云展 动力","https:\u002F\u002Fimages6396289.oss-accelerate.aliyuncs.com\u002F2299021137\u002Fow873ffgpa\u002F鸿威宣传视频.mp4","https:\u002F\u002Fimages6396289.oss-accelerate.aliyuncs.com\u002F2299021137\u002Fbx9hvodhys\u002F广州本部.jpg","https:\u002F\u002Fwww.ysdli.com\u002Fspc.html?m=luNgqWJZN","1405769664983003136","c99s2fxwoa6zj",-2,null]