广告
加载中

从Human到Humagent:企业AI组织转型的思考

赵艳秋 2026-06-18 22:27
赵艳秋 2026/06/18 22:27

邦小白快读

EN
全文速览

本文围绕AI时代的产业变革,梳理了核心信息和普通人可参考的应对干货,具体如下:

1.核心判断:AI不会完全替代人类,人机协作会长期存在,未来企业组织会是Human加Agent的Humagent复合模式,分工上AI承担繁重、高效的执行类工作,人类发挥前瞻性优势,负责把控价值方向、做判断决策,二者是伙伴而非对手,不用过度恐慌被替代。

2.个人应对方向:当前企业AI转型中,对于可被AI替代岗位的员工,企业会将其转向需求管理、架构设计等更能发挥人类优势的岗位,而非直接裁员。普通人需要主动改变认知,避免排斥AI或轻视AI两种极端,主动学习拥抱AI,学会和AI协作创造更大价值。

本文为AI时代品牌商的AI转型和业务布局,梳理了干货内容,具体如下:

1.核心趋势判断:AI已经颠覆了产业底层逻辑,不主动转型就会被淘汰,品牌商做AI转型不能只把AI当工具,核心是要推动组织整体进化,把AI Agent当做数字员工纳入组织才能真正提升整体能力。

2.业务端机会:AI Agent可以完成从市场调研、产品设计到成本分析的端到端工作,效率远高于传统模式,能大幅提升产品研发效率,品牌商可重构研发流程,通过风险审核机制管控AI出错的问题,发挥AI的效率优势。

3.转型实操原则:转型要遵循10-20-70原则,10%投入算法,20%投入数据技术,70%投入人员、流程和文化转型,转型是一把手工程,要做好文化培训,调整KPI导向鼓励人机协作,做好数据治理保障AI应用合规可控。

本文梳理了AI变革给卖家带来的机会、风险与转型指引,干货内容如下:

1.风险提示:AI已经带来产业颠覆性变化,AI coding可实现效率提升上百倍,不主动转型的卖家很难在竞争中胜出;当前Token供需严重失衡,需求指数级增长,供给仅能年增20%-30%,今明两年算力都会是转型瓶颈,基础设施建设还存在建成即过时的悖论,需要提前应对。

2.机会提示:人机协作为长期趋势,卖家可将AI Agent作为数字员工重构组织,大幅提升运营、研发效率,打开新增长空间;卖家可根据自身情况选择Token方案,中小卖家无需自建,有大量隐私数据的中大规模卖家可选择自建平衡成本与安全。

3.转型参考:转型要一把手牵头,7成投入放在组织、人员、流程变革上,建立AI风险管控机制,调整人员到适配岗位,做好数据治理保障合规。

本文对AI时代工厂推进转型、抓住变革机会给出了明确启示,干货内容如下:

1.生产设计端变化:AI Agent可以完成从市场调研、产品设计到成本分析的全流程工作,知识宽度和效率远超人类,工厂可引入AI Agent作为数字员工,重构研发生产流程,大幅缩短新品周期,提升效率,同时要建立分阶段的风险评审机制,管控AI出错的风险。

2.组织转型启示:工厂做AI转型不能只把AI当工具,要推动组织整体进化,遵循10-20-70投入原则,70%的投入要放在人员、流程和文化转型上,转型需要一把手牵头推动,调整原有组织架构适配新模式。

3.机会与基础设施启示:当前AI产业处于爆发期,工厂有大量隐私生产数据可自建Token工厂平衡安全与成本,要敢于超前投资AI基础设施,分级利用算力,先进算力做训练,存量算力做推理,还可联合上下游开展软硬件联合创新提升整体效率。

本文梳理了当前企业AI转型的行业趋势、核心痛点,给出了服务商可参考的发展方向,干货内容如下:

1.行业发展趋势:AI已经改变了生产力核心要素,传统企业AI转型需求全面爆发,未来主流组织形态会是Humagent即人类加AI智能体的复合组织,AI基础设施是转型核心刚需,上下游协作共同转型是长期发展方向。

2.客户核心痛点:70%的企业AI转型阻力来自非技术因素,多数企业只把AI当工具,个体AI能力提升但组织能力跟不上,组织力成为最大转型摩擦力;同时当前Token供需失衡,今明两年算力紧张,企业建设AI基础设施还面临规划不足要推倒重来、建成即过时的悖论。

3.业务方向参考:服务商可转型定位为企业AI转型的同行者,而非仅售卖AI基础设施,要为客户提供Humagent转型的顶层设计、方法论、流程重构配套服务,还要推动软硬件联合创新,帮助客户提升同等算力的Token产出,缓解算力压力降低成本。

本文围绕企业AI转型的核心需求,为AI平台商的运营发展给出了干货参考,内容如下:

1.企业对AI平台的核心需求:企业做Humagent转型,需要平台支撑新型组织的运行,需要适配组织转型的基础设施、数据治理工具、AI运行管控能力,需要平台支持建设可追溯、合规可控的超级数据空间,同时需要平台帮助降低Token成本,缓解当前算力紧张的问题。

2.平台可参考的实践做法:平台可调整定位,从单纯卖AI基础设施转向做企业AI转型的同行者,除基础设施外还要输出组织转型的方法论,帮助客户完成顶层设计与流程变革;可开放自有Token资源支持企业内部创新,推动上下游联合开展软硬件创新,提升单位算力的Token产出。

3.风险规避方向:平台要提前预判Token需求的指数级增长,提前布局算力,引导客户分级利用算力,还要帮助客户建立AI风险管控机制,防范数据安全与AI误操作风险,保障AI运行合规可控。

本文提出了AI组织转型的新概念与一线实践,总结了AI产业的新动向新问题,对产业研究有较高参考价值,内容如下:

1.产业新动向:当前AI已经撼动了知识社会的底层逻辑,改变了生产力核心要素,人机协作是产业长期发展方向,产业界已经开始探索Humagent即人类加AI智能体的新型企业组织形态,AI转型中组织变革的重要性已经成为产业共识,软硬件联合创新成为突破算力瓶颈的新方向。

2.待解决的新问题:AI发展带来了人类价值重新界定的伦理问题,企业AI转型中组织力是最大阻力,7成转型阻力来自非技术因素;当前Token供需严重错配,今明两年都会面临算力紧张,企业建设AI基础设施存在规划不足和建成即过时的悖论,Humagent转型目前没有成熟的路径与方法论,仍处于摸着石头过河的探索阶段。

3.新商业模式启示:AI服务商已经出现从卖AI基础设施的卖铲人,转型为企业AI转型同行者的新商业模式,不同规模企业选择租用或自建Token会最终形成商业平衡,为产业研究提供了全新的实践样本。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

This article distills key insights on industrial transformation in the age of AI and shares actionable takeaways for the general public:

1. Core takeaway: AI will not fully replace humans. Human-AI collaboration will remain the norm long-term, and future organizations will adopt a "Humagent" hybrid model combining human workers and AI agents. Under this model, AI takes on repetitive, high-volume execution work, while humans leverage their strengths in foresight to set value directions and make strategic decisions. The two sides are partners, not rivals, so there is no need for excessive fear of job displacement.

2. Personal action steps: During the current wave of corporate AI transformation, companies will reassign employees in AI-replaceable roles to positions that play to human strengths, such as demand management and architectural design, rather than implementing direct layoffs. Ordinary people should actively update their mindsets, avoid the two extremes of rejecting AI or underestimating it, proactively learn to work with AI, and master collaboration with AI to create greater value.

This article shares key insights and actionable guidance for brands pursuing AI transformation and business布局 in the AI era:

1. Core trend outlook: AI has already upended the underlying logic of industries. Brands that do not proactively transform will be displaced. For brands, AI transformation is not just about adopting AI as a tool—its core is driving overall organizational evolution. The real performance gain comes from integrating AI agents into the organization as digital workers.

2. Business opportunities: AI agents can complete end-to-end work from market research and product design to cost analysis, with far higher efficiency than traditional workflows, dramatically accelerating product R&D. Brands can restructure their R&D processes and leverage risk review mechanisms to mitigate AI errors, unlocking AI’s efficiency advantages.

3. Practical transformation principles: Transformation should follow the 10-20-70 rule: allocate 10% of investment to algorithms, 20% to data technology, and 70% to personnel, process and cultural transformation. AI transformation is a top leadership initiative; brands must deliver targeted cultural training, adjust KPI frameworks to incentivize human-AI collaboration, and implement robust data governance to keep AI applications compliant and controllable.

This article outlines the opportunities, risks, and transformation guidance brought by AI disruption for sellers:

1. Risk alert: AI has already brought disruptive industrial change. AI-powered coding can boost efficiency by 100x, leaving sellers that refuse to transform at a severe competitive disadvantage. Currently, there is a severe supply-demand imbalance for AI tokens: demand is growing exponentially, while supply can only increase by 20-30% annually. Computing power will remain a bottleneck for transformation over the next two years, and AI infrastructure also faces the "built-in-obsolete" paradox, requiring proactive preparation.

2. Opportunity outlook: Human-AI collaboration is the long-term trend. Sellers can restructure their organizations by adopting AI agents as digital workers to dramatically boost operational and R&D efficiency, unlocking new growth avenues. Sellers can choose a token strategy aligned with their scale: small and medium-sized sellers do not need to build their own systems, while mid-to-large sellers with large volumes of sensitive private data can build in-house systems to balance cost and security.

3. Transformation guidance: Transformation must be led by top leadership, with 70% of investment allocated to organizational, personnel and process changes. Sellers should also establish an AI risk control framework, reallocate staff to roles suited to the new model, and implement data governance to ensure compliance.

This article shares clear insights for factories to advance AI transformation and capture opportunities from industrial change:

1. Changes in production and design: AI agents can complete end-to-end work from market research and product design to cost analysis, with far greater knowledge breadth and efficiency than human teams. Factories can introduce AI agents as digital workers to restructure R&D and production processes, drastically shorten new product development cycles, and boost efficiency, while establishing a phased risk review mechanism to mitigate the risk of AI errors.

2. Organizational transformation insights: AI transformation for factories is not just about adopting AI as a tool—it requires driving overall organizational evolution, following the 10-20-70 investment rule, with 70% of investment allocated to personnel, process and cultural transformation. Transformation must be led by top leadership, and existing organizational structures need to be adjusted to fit the new model.

3. Opportunities and infrastructure insights: The AI industry is currently in a period of explosive growth. Factories holding large volumes of private production data can build in-house token facilities to balance security and cost. Factories should dare to make forward-looking investments in AI infrastructure, tier computing power usage (allocate advanced computing power for model training and existing computing power for inference), and collaborate with upstream and downstream partners on joint software and hardware innovation to improve overall efficiency.

This article sorts out industry trends and core pain points in current enterprise AI transformation, and outlines actionable development directions for service providers:

1. Industry trends: AI has already reshaped the core factors of productivity, and demand for AI transformation from traditional enterprises is exploding across the board. The dominant future organizational model will be "Humagent", a hybrid structure of human workers and AI agents. AI infrastructure is a core刚需 for transformation, and joint transformation across upstream and downstream value chains is the long-term development direction.

2. Core customer pain points: 70% of resistance to enterprise AI transformation comes from non-technical factors. Most enterprises only treat AI as a tool, leading to a mismatch between improved individual AI capabilities and lagging organizational capabilities, making organizational inertia the biggest bottleneck to transformation. Meanwhile, the current token supply-demand imbalance will keep computing power constrained over the next two years, and enterprises also face the paradox of inadequate planning leading to reworks and infrastructure becoming obsolete immediately after completion.

3. Recommended business directions: Service providers should reposition themselves as partners for enterprise AI transformation, rather than just sellers of AI infrastructure. They should offer top-level design, methodology, and process restructuring services to support clients’ Humagent transformation, drive joint software and hardware innovation, and help clients increase token output per unit of computing power to alleviate computing pressure and reduce costs.

This article shares actionable guidance for AI platform operators based on core demands from enterprises undergoing AI transformation:

1. Core enterprise demands for AI platforms: For enterprises pursuing Humagent transformation, platforms need to support the operation of new organizational models, provide infrastructure, data governance tools, and AI operation control capabilities adapted to organizational transformation, support the construction of traceable, compliant and controllable super data spaces, and help enterprises reduce token costs to ease the current computing power shortage.

2. Recommended operational practices: Platforms should adjust their positioning from pure AI infrastructure sellers to partners for enterprise AI transformation. In addition to infrastructure, they should also output methodology for organizational transformation to help clients complete top-level design and process restructuring. They can open their own token resources to support internal innovation for client enterprises, and drive joint software and hardware innovation across upstream and downstream to increase token output per unit of computing power.

3. Risk mitigation guidance: Platforms should proactively prepare for exponential growth in token demand, pre-deploy computing power resources, guide clients to tier computing power usage, and help clients establish AI risk control mechanisms to prevent data security risks and AI operational errors, ensuring AI operation remains compliant and controllable.

This article puts forward new concepts and frontline practices for AI-enabled organizational transformation, summarizes new trends and emerging problems in the AI industry, and offers high reference value for industrial research:

1. New industry trends: AI has already shaken the underlying logic of the knowledge society and reshaped the core factors of productivity. Human-AI collaboration is the long-term development direction for the industry. Industry players have already started exploring "Humagent", a new organizational form combining human workers and AI agents. The importance of organizational change in AI transformation has become an industry consensus, and joint software and hardware innovation has emerged as a new direction to break through computing power bottlenecks.

2. Unresolved emerging issues: AI development has raised ethical questions around redefining the value of human work. Organizational capability is the biggest barrier to enterprise AI transformation, with 70% of transformation resistance coming from non-technical factors. Currently, there is a severe supply-demand mismatch for tokens, and computing power will remain constrained over the next two years. Enterprises building AI infrastructure face the paradox of inadequate planning and infrastructure becoming obsolete immediately after completion. No mature path or methodology exists yet for Humagent transformation, which remains in an exploratory stage.

3. Insights for new business models: AI service providers have begun shifting from the traditional "pickaxe seller" model of selling AI infrastructure to a new business model as transformation partners for enterprises. Different-sized enterprises’ choices between renting tokens and building in-house systems will eventually reach a commercial balance, providing a brand-new practical sample for industrial research.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

人机协作将会长期存在。人的优势是什么?AI的价值又是什么?

文|赵艳秋

编|牛慧

人对AI的态度,正变得复杂而矛盾。就在刚刚过去的5月,美国各大高校毕业典礼接连发生同类事件:只要嘉宾一味推崇AI、淡化就业冲击,台下毕业生就集体喝倒彩。年轻人用最直接的方式表达自身面临“被替代”的惶恐与不满,也道出在社会评价体系中人类贡献该如何重新界定的迷茫。

“前几次产业革命,人的地位没有被撼动,但AI撼动了人存在的基本价值,并由此触及伦理、道德与文明存续这样的底层问题。”在6月16日举办的人工智能+生态大会(AIEC2026)上,浪潮信息董事长彭震说,知识社会最底层的逻辑开始发生巨变,这可能对千行百业带来颠覆性,未来变化还难以预知。

“实际上,我们看到AI的优点很明显,但缺陷也很突出。”彭震说,“为此,我们认为人机协作将会长期存在。在这个过程中,我们不是思考AI是否要替代人类,而是思考人类如何与AI和谐共处,如何利用AI创造更大价值。”彭震强调,人类需要学会与AI相处。他进而提出一个新概念——Humagent,即未来企业组织应该是Human加Agent的复合体。而企业也将面向Humagent展开全方位变革。

01

企业AI转型,需要组织进化

很多人将AI视作一种工具,但彭震指出,AI与以往不同。数字化、互联网化更多是工具的改变,如网约车解决了空间、效率和执行力问题,但人始终是主体。“但AI拥有更强的智慧甚至超越人类,它的核心是改变了劳动者和人,这是生产力中最关键的要素。”

而这场变革无法阻挡。根据IDC数据,传统SaaS企业5年后市占率仅剩1%;AI coding已让软件公司估值大幅下挫、规模裁员,效率提升上百倍。“你与一个效率比你高100多倍的对手竞争,能赢吗?”彭震说,如果不主动改变,就要被改变,而那个结局将极其悲惨。

不过,当很多企业砸钱上了AI,却发生了一个现象:个体能力提升了,组织能力却没提升。彭震认为,根源在于企业对Agent的定位。“我们到底把AI看成什么?如果只把它当成智能化个人助理或简单工具,它发挥的作用有限;如果把它视为数字员工、企业的一份子,情况就会不同。”简言之,把Agent当组织成员,有了组织进化,企业能力才能提升。

这也解释了一个反差,AI转型中,最大的阻力不是技术。“组织力成为最大摩擦力。”彭震引用第三方数据,企业AI转型中非技术因素占比高达七成。“模型能力可以快速提升,但组织能力跟不上,业务效果就不理想。”由此,他提出“10-20-70” 原则:10%投入算法,20%投入数据和技术,70%投向人员、流程和文化转型。

那么,组织究竟要怎么变?

第一重变化,是协作单元被重写。彭震认为,Agent的优势在于知识宽度远超人类,且效率极高。过去企业有那么多工种,是因为人受限于知识密度、只能专精在某一领域,企业要把不同专业的人串行或并行,才能产生确定性的商业结果。而今天从市场调研、架构设计到底层代码、成本分析,Agent可以端到端完成。为此,需要给Agent更大的数据集,做流程再造。但Agent也有明显短板,组织再造要“风险防范、成本防范,分阶段、分专业领域评审”,更要管住AI,“要不它今天可能把电闸拉了,明天把房子拆了”。

“我认为企业效率提升的前提,是将Agent从个人助理变成企业数字员工,依赖组织进化,支持围绕Humagent新组织模式的变革。”彭震说,既要发挥Agent的价值智力创造,又要发挥人在长期训练中积累的稳健性和可靠性。“扬长避短,形成类似于超级团队的模式。”

那么在超级团队中,人要发挥什么优势?清华大学全球产业研究院院长彭凯平认为,人的优势是前瞻性。“我们大脑有一个默认神经网络,就是下意识憧憬未来,不断构建未来的可能性。”在他看来,人与AI的分工本质类似“道与术”:智能体扛下繁重的执行,把任务做到极致;人则把控价值和方向,做判断、做决策、赋予意义。“人与AI不是对手而是伙伴。”

第二重变化,是转型必须由一把手扛起。既然要将Agent当数字员工,AI转型就成了绝对的“一把手工程”,需要领导者根本性的思维转变。此前不少企业把AI当工具,交给IT部门推动,只敢部署在不出错但价值较低的场景,如会议纪要、人力资源助手。但AI是要改变企业经营结果、创新业务、优化供应链、反欺诈。

在浪潮信息,由彭震牵头,IT部门改名为“智能化转型部”,抽调各部门业务专家与AI专家,负责顶层设计,数据治理与KPI设立。彭震把这个新角色称为公司CAIO(Chief of AI Office),“我们对他的要求不是编程,核心是重构公司整个AI的顶层设计,业务创新则交给业务部门去干。”

这一判断正成为产业领导者的共识。百度董事长李彦宏认为,组织正从“人与人分工”进化为“人机混合编队”;李开复则预判将出现新组织形态:一个人对某项跨职能结果端到端负责,围绕他协同的是Agent专业化集群,而“如果AI部署没有改变任何一个财报电话会上的数字,那就不是转型,只是建了个昂贵的AI实验室”。

02

围绕Humagent,一场AI原生探索

提出Humagent新组织模式的同时,浪潮信息已动身探索。

第一道坎是人的改变。“企业AI转型,最大的摩擦力是人。”彭震观察,员工对AI容易产生两种极端认知:一种出于生存本能,对使用Agent天然排斥,另一种是有老员工怀疑Agent是否可行。彭震认为,转型的第一件事是文化变革,给文化先松土。浪潮信息对全员开展了13门理论课的培训考试,让大家认识、拥抱Agent,避免要么惧怕它,要么鄙视不用它。

松土的过程中,他们看到年轻人在创新上的作用令人惊叹。公司举办黑松客大赛,在其中找到更有意愿改变、有能力的年轻人。“这种内驱力是企业变革中最需要、最紧缺的资源。有时技能可以训练,但内驱力培养确实很难。”

更关键的是导向。“我们现在采用Humagent的观点,不是谁取代谁或彼此PK,我们使用KPI、定义KPI,更多告诉员工如何和Agent一起创造更大价值。”彭震说。即便部分Coding和测试岗位可被AI替代,人将被转向需求管理、架构设计或帮助业务部门做Agent转型,不是裁员,而是去更能发挥人类员工优势的岗位。

那么,如何让Agent更稳妥的落地?浪潮信息正在重构许多研发流程。最典型的流程重构是AI Coding,有业界判断,编程已占模型Token消耗量的九成以上。但大家很快发现,Agent代码产出率很高,却经常犯错,甚至是常识性错误,“这就像年轻员工干劲很足,但经验不足”。彭震提出一个思路,把它当“新员工”来管。“我们需要思考如何管理和激发‘新员工’的作用,又不被他们带偏,这要更多依赖企业严格的风险防范机制和流程机制。”

彭震介绍,过去从代码需求解读到发布全流程,已形成一套面向人的管理流程体系,在Agent时代这套体系依然要发挥作用。由于采取了多种纠正问题的措施,总产出稳定可靠,符合商业标准。

Humagent不只涉及人,还包含Agent,因此整套流程需要重建,否则它的节奏和能力完全发挥不出来。如彭震所说“过去组织运转靠的是个人经验的积累转化为企业的标准流程和动作。在Agent时代,Agent靠什么来积累经验?是数据。”因此,所有决策和过程数据必须保存,可追溯。彭震强调,企业要建设一个面向Humagent的超级数据空间。浪潮信息已从公司整体数据治理层面做顶层设计,避免数据碎片化和泄露。尤其关键的是,不能让AI直接操作原数据,AI空间的数据若要回写,也必须经业务流程管理和人工把关,确保合规、可管、可控。在浪潮信息,这一切跑在自研的元脑EPAI平台上。

彭震坦言,今天围绕Humagent的转型,最大的困难是路径、顶层设计、方法论、工具都没有先进经验,“要摸着石头过河”。在此过程中,浪潮信息走进美的、小米和一众互联网公司,看一手资料,“自己想、自己变,你自己就是活生生的例子”。与此同时,浪潮信息重新定位了自己:从卖AI基础设施的“卖铲人”,转向企业AI转型路上的“同行者”。

03

Token紧俏,企业AI基础设施怎么选怎么建

AI转型涉及组织进化,同时也需要AI基础设施的支撑。

今年以来,算力变得越来越紧张,优质token变得极为紧俏。彭震告诉数智前线,其中的原因是供需严重不匹配:Token需求一年增长几十倍,指数级增长,但算力供给,包括芯片、存储器、光纤的产能增长却是线性的,扩展速度每年最多20%~30%。“我们认为今明两年都很困难,这是令人焦虑的结论。”从某种角度来看,基础设施可能会成为瓶颈。为此,很多互联网企业、大模型公司已开始大量淘汰低质量Token,转向能力更强但更贵的模型。

企业选择怎样的AI基础设施?彭震认为这里有复杂的考量因素,并非每个企业都要建立token工厂,例如一人公司。有些企业类似浪潮信息、美的,内部有很多隐私数据,必须建立token工厂。同时,最近token价格变贵,也催生了“反制力”。彭震观察到一场博弈正在发生:当租用云上Token的成本高到一定程度,企业就会转向自建。“这是一个复杂的商业博弈,最后大家会形成一个平衡点。”

在Token昂贵的情况下,浪潮信息内部特意留出一些免费、自建的Token给员工做创新。当企业内部创新真正进入生产时,需要很高的费用,“如果因为token贵,用不起,员工的翅膀就被绑起来了,我们自建Token,也是给创新的一种松绑。”

彭震认为,AI时代基础设施有时具有决定性作用。“信息化时代,基础设施更多承载一些业务,要求并不高。而AI时代基础设施能力,从某种程度可能决定企业的智慧能力。智能涌现依赖于算力、大模型规模、数据量以及处理效率。”

不过,如今企业在建设AI基础设施时,确实面临一些挑战:一方面投资需要具备前瞻性和预见性,浪潮信息遇到很多客户,启动建设时发现原有规划已经不够,要推倒重来;但另一方面,大家又面对一个悖论,即基础设施技术迭代非常迅速,建成即过时。不过,彭震看到,现在企业往往将最先进的设施用于训练,稍有差距的用于推理,不断循环往复。“今天算力紧张的情况下,我们看到3年前的算力一样发挥了巨大价值。”面对Token的极速增长,有时在基础设施方面,企业确实要敢于投资,也要超前投资。

在他看来,长期来看Token价格一定会降低,目前业界有两个新动向:一是所有大模型公司都在投入做一件事——通过技术改善,让自己在同样算力上产生远远大于竞争对手的Token。另一方面,在基础设施领域,过去大家更多沿着摩尔定律方式,但最近一个进展是软硬件的联合创新。“我们与国内企业进行联合创新,他们提出新算法,我们采用新介质、硬件及网络方式适配算法,带来的收益非常大。”美国企业也在开展类似创新,越来越多的人在研究这一新路径。

而彭震也强调,AI转型不能只靠一家企业——更多还要依赖产业界的上下游,共同面对新一次产业变革。

注:文/赵艳秋,文章来源:数智前线(公众号ID:MzkwNDMyOTA1NA==),本文为作者独立观点,不代表亿邦动力立场。

文章来源:数智前线

广告
微信
朋友圈

这么好看,分享一下?

朋友圈 分享

APP内打开

+1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0