2021，百度飞桨交出最新成绩单

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"12月12日，在2021深度学习开发者大会现场，百度首席技术官、深度学习技术及应用国家工程实验室主任王海峰公布了百度深度学习框架飞桨的最新成绩单："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"截止目前，飞桨已经凝聚了406万开发者，创建了47.6万模型，服务于15.7万企事业单位。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"发布十大最新技术和生态进展"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"大会现场，飞桨发布十大最新技术和生态进展，包括飞桨新版全景图—产业级模型库新增文心大模型、业界首个产业实践范例库、飞桨“大航海”2.0共创计划、飞桨开源框架v2.2 体系化新增科学计算API、端到端自适应大规模分布式训练技术、文本任务全流程加速、多层次低成本的硬件适配方案、产业级开源模型库模型超过400个、企业版升级自动高效的模型部署功能，以及1分钟极速安装完成本地高效建模的飞桨EasyDL桌面版。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/infoq\/e2\/e2cdab98f1decfe56f7157a0798c9526.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"具体来说，技术层面，飞桨全新发布的开源框架v2.2，具备四大特性：飞桨API更加丰富、高效、兼容，新增大量科学计算API；高效支持超大模型训练的端到端自适应大规模分布式训练技术；全流程加速文本任务，解决文本领域开发在性能和训推一体方面的痛点问题；多层次、低成本的硬件适配方案，降低框架与芯片的适配成本。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/infoq\/7b\/7b53e0911220c0095e79e06ca83b49fb.png","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"飞桨产业级模型库新增百度最新发布的知识增强文心大模型，让大模型进入产业应用；官方支持的产业级开源算法模型超过400个，并发布13个PP系列模型，在精度和性能上达到平衡，将推理部署工具链打通。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"产业落地方面，飞桨推出了业界首个产业实践范例库，从真实产业场景分析、完整代码实现，到详细过程解析，直达项目落地，覆盖数十个高频应用场景，推动AI落地可复制和规模化。面向产业场景提升开发效率和资源使用效能的飞桨企业版升级了自动高效的模型部署功能，同时推出1分钟极速安装完成本地高效建模的飞桨EasyDL桌面版。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"生态方面，飞桨全新发布“大航海”计划2.0，在启航、护航、领航三大航道基础上，新增“共创”计划，以飞桨平台为基座，社区开发者共创工具、模型、产业案例与实践经验；形成产业创新需求对接平台，共创产学研用正循环；与生态伙伴一起建设人工智能产业赋能中心，共创区域创新生态。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"百度集团副总裁、深度学习技术及应用国家工程实验室副主任吴甜表示，全国各城市的AI开发者规模逐年增长，应用人工智能的企业数量，也呈现了多地开花、多行业繁荣的景象。从推荐等通用场景，到客服系统智能派单等行业衍生场景，再到发电预测等行业关键场景……人工智能与产业场景的结合越来越深入、越来越专业。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"王海峰：人工智能工业大生产阶段，具备融合创新和降低门槛两大特点"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"深度学习框架下接芯片，上承应用，相当于智能时代的操作系统，对于AI技术发展和产业落地至关重要。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“深度学习推动人工智能进入工业大生产阶段。面向技术和产业发展需求的AI大生产平台可以让AI技术以标准化、自动化和模块化的方式输出给千行百业，实现规模化应用，同时以平台为基础促进融合创新、共同发展。飞桨产业级深度学习开源开放平台是典型的AI大生产平台，赋能广大开发者，有力支撑AI工业大生产，促进技术创新和产业智能化升级。”王海峰表示。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"王海峰认为，目前，人工智能呈现出“融合创新”和“降低门槛”的特点："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"一方面，AI技术及产业的融合创新越来越多。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"融合创新包括知识与深度学习融合、跨模态多技术融合、软硬一体融合、技术与场景融合。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"知识与深度学习融合方面，百度刚刚发布的全球首个知识增强千亿大模型—鹏城-百度·文心，参数规模达到2600亿，在60多项任务中取得最好效果。知识增强千亿大模型的发布，也得益于飞桨的大规模训练技术。飞桨研制了自适应大规模分布式训练技术，使大模型训练效率达到业界最好水平。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"跨模态多技术融合方面，机器需要像人类一样理解真实世界，听懂语音、看懂图像视频、理解语言。百度研制了知识增强跨模态大模型，提升跨模态理解与生成能力。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"软硬一体融合方面，芯片和深度学习框架综合算力、功耗、时延等因素联合优化，才能获得AI应用的最佳效果。飞桨已经和22家国内外硬件厂商完成了31款芯片的适配和联合优化工作。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"技术与场景融合方面，飞桨在应用于各行各业的过程中，与实际应用场景融合创新，既切实解决了行业应用问题，又使飞桨平台得到持续积累和提升。飞桨还专门推出系列科学计算API，支持量子计算、生命科学等应用，探索AI与前沿科技的融合创新。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"另一方面，虽然AI技术越来越复杂，但AI开发与应用的门槛却越来越低。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"技术方面，近期成为AI技术重要方向的大模型，具有通用性好、泛化性强、效果好等特点，可降低AI开发与应用门槛。飞桨在核心框架上实现的动静统一，兼顾科研开发的灵活和产业开发的高效，可满足不同类型开发者需要。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"人工智能人才培养的机遇和挑战"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"人工智能行业，无论是技术发展，还是产业升级，归根结底要靠人才实现。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"吴甜现场分享了百度在与吉林大学携手飞桨探索产教融合协同育人路径，实现产学研用正循环的实例。吉大三创（创意、创新、创业）实验室团队针对医药企业药瓶缺陷检测场景需求，基于飞桨平台创新研发了一套全方位、高精度的自动化检测系统，并实现了产线落地。吉大与飞桨共建了CV及AI+质检课程体系，将课程资源全面开放，通过飞桨学习与实训社区AI Studio，将这一实践经验输送给百万开发者，实现了产学研用正循环。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"王海峰认为，应该培养复合型AI人才。随着AI与产业的深度融合，需要越来越多既懂AI，又具备产业经验的复合型AI人才。一直以来，飞桨致力于复合型AI人才培养，联合学术界和产业界，打通产学研用的正向循环，持续为社会贡献AI人才。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在大会圆桌论坛环节，百度研究院大数据实验室和商业智能实验室主任、美国俄勒冈大学计算机系教授窦德景，复旦大学特聘教授、原复旦大学计算机科学技术学院、软件学院院长王晓阳，上海大学计算机工程与科学学院院长谢少荣，华东师范大学计算机科学与技术学院副院长、华东师范大学上海智能教育研究院副院长周爱民等多位在人工智能人才培养领域中有深入研究，在产教融合的道路上有深刻见解的教育专家，共同探讨了现阶段，AI人才培养的机遇与挑战。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"AI人才培养，如何实现产教融合？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"窦德景：人工智能和产业结合日益紧密，从高校的角度来看，企业应该如何发挥技术和产业的实践优势，助力高校人才培养实现产教融合？"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"周爱民："},{"type":"text","text":" 企业和高校各自职责不同。对于高校来说，特别是对于研究生和博士阶段的人才培养来说，专业硕士、非全日制硕士培养方面能够和企业形成良好互动，企业可以建立相关的实习基地，可以邀请企业导师发挥各自的优势。高校在人才培养上有更加全面的知识体系，覆盖面更全，特别是人工智能为代表的新工科的专业实践性更强，对人才的需求非常大，因此在这一块可以发挥高校和企业界的优势联合来培养人才。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"王晓阳："},{"type":"text","text":" 业界常有人发问，为何高校教育出来的人与产业界的应用总是脱节。这其中的一个原因是，高校教育和产业的运行节奏是不同的，需要相互配合。谈到企业如何助力高校人才培养，第一企业应该要有耐心，很多时候不能说一种新技术出来了，就单纯觉得学校培养的人不适合企业发展。第二，应该探索出一种共同育人的模式。从很早开始，企业希望影响学校教学，除了在专业、学位上更深入外，在通用技术方面也要多做努力。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"通用技术较容易在学校推动。很多企业想推广产品，让学生使用它的产品，培养潜在的用户，其实这有点太急功近利。应该把着眼点放在如何把更通用的技术跟学校的教育结合起来。因为整个计算机科学比较深，在哪里区分，在某些专业推行哪一些技术，需要做研究，需要长期耕耘。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"谢少荣："},{"type":"text","text":" 高校对本科生有系统知识的培养。特别是高年级以后的学生，有了共性的专业基础后，要注意跟应用结合。在这时，高校学生培养希望跟企业的一些很新的、很先进的算法模型，开源平台等结合起来。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"曾经有一个院士讲，高校里面不要用昨天的知识教今天的学生，让他们去面对明天的需求。所以我们始终在思考这个问题，也在做各种探索，产教融合应该把最新研发的东西及时纳入到专业培养里面来，这是很有必要的一个环节。也期望像百度这样的企业多给予高等教育支持和帮助，共同育人，不止是算法和平台，还包括一些研发工程师甚至可以走进课堂、走进高校，一起来培养学生，培养未来的人工智能人才。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"现阶段的高校课程设置要调整吗？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"窦德景：产业发展新阶段，国内高校各有特色，侧重点不一样，但所有的高校都得教课，咱们在制定教学大纲的时候，编程能力和AI的基础知识，应该做相应的调整吗？"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"谢少荣："},{"type":"text","text":" 上海大学设计人工智能专业培养体系的时候进行了很多调研、分析和思考，我们的培养理念是，首先让学生打好基础，在数学模块、专业基础模块的基础必须打牢打扎实。我们针对专业课分了六大模块，要有基础理论人才培养类型，也要有AI+复合型人才的培养类型，其中有两个模块是偏向于通用公共基础理论方面。我自己是搞智能无人系统的，其中有一个模块就是结合智能无人系统来进行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"王晓阳："},{"type":"text","text":" 计算机行业，从最底层到抽象层、应用层，分了很多层面，从横向的切成很基础的算法、编程语言的问题，以及最上层怎么应用的问题。数字还要分成各种各样的专业，像无人驾驶和自然语言分析就很不一样，可以有纵向的划分，在这个层面上，计算机专业本身确实跟现在整个社会对计算机科学的要求有点脱节，专业的分置和布置都在变。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"刚才我讲到为什么节奏不太一样？高校的专业一设置就是四年，学生一般进来这个专业，四年不大可能变了。这四年对于产业界来讲是很长的时间，课程设置既要对目前产业发展、社会发展的需求做出响应，另一方面又要维持它的长远性，这两个需求永远在互相竞争。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"目前看来，我觉得要在在横向选择时要有不同的侧重点，如侧重不同的学校以及不同特色的学校。很多企业对行业比较了解，对于各种各样的行业应用需求比较了解，企业可以帮助学校进行专业设置等。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"周爱民："},{"type":"text","text":" 我觉得要发挥各自学校的优势和特色。华东师范大学最大的特色是，我们是一个师范类见长的综合性大学，所以我们在想办法来突出师范的特色。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"去年我们在计算机科学与技术一级学科下，自主申报了智能教育二级学科的博士点，对于这样的学生，我们在课程体系的设置上体现了交叉的特性，首先学生需要知道计算机和人工智能的技术作为基础，基本计算机的基础有相应的模块。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"还有一个很重要的模块就是教育，因为是交叉的，而且不是理科和工科的交叉，而是文科和工科的交叉。对博士班的培养方案，除了基本的课程学习以外，另外一块内容是有关智能教育这方面科学技术的研究，要深入地参与到上海智能教育研究院的课程中。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们不仅仅想要做简单的科学研究，还想把做科学研究的成果最终落地到中小学课程教学过程中。所以第三个模块就是一定要实践，把这些科学研究的理论成果参与到相关产品的开发过程中，最终从早期的理论学习到科学的研究到产业的落地，形成一个完整的链条。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"人工智能人才培养要结合区域特色"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"窦德景：中国各个地方都有自己的特色，例如长三角就是一个不论在经济还是科技上都非常有活力的地方。结合到人工智能人才的培养，一个问题是，怎么样和当地的产业（企业）相结合？"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"王晓阳："},{"type":"text","text":" 照理说，计算机科学、计算机应用，是一个普适性学科，但到了应用层面，各个区域应用的专注点不太一样。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"今天人工智能的应用很大程度上不光是计算机专业的问题，区域的特殊性很大程度上反映了学校的特殊性。像我们复旦大学经济、金融相当发达，在人工智能应用方面有很大的潜力，我们现在推的这些计算机工具、人工智能工具越来越轻盈、越来越好用，好用不光是为了计算机从业者建模更容易，更是希望对应用方来说落地更容易。如何在不同的区域推动不同专业领域使用人工智能的工具，这是一个值得思考的问题。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"复旦是一直讲通识教育比较多的学校，理念是通识教育为主，专业教育在后，我们提出，通识教育里面应该有计算机、人工智能。专业的人士，比如学经济的应该搞一个人工智能的小专业，这样使得他们在使用人工智能工具的时候，在区域发展的过程当中能够发挥比较好的作用，我觉得这块大有发展潜力。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"复杂的国际环境下，如何培养AI人才"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"窦德景：当下国际环境很复杂，在这样的背景下，如何做好AI人才培养？"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"王晓阳："},{"type":"text","text":" 这个问题比较复杂，我们要在应用人才的培养上做一些工作，但在当下复杂的国际环境下，我觉得最重要的一点是在基础研究上加大投入。像百度这样的企业，我希望能够在基础演讲上更加给力。如何做好基础研究，希望高校和企业增加合作，共同探索更好的途径。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong"}],"text":"谢少荣：在"},{"type":"text","text":"这样的国际形势下，我们中国人，一直在提科技的自立自强。而实现技术自立自强，人才，特别是未来的人才非常重要。希望用我们自己的算法、模型、开源平台，来培养我们自己未来的AI人才，这非常重要。所以我很期望像百度这样的大企业，企业有资源和实力，我们学校也有开放的心态，期望和大家一起联合起来，培养好未来的人，人是最重要的。"}]}]}