广告
加载中

雷军重仓一位31岁女将

刘博 2026-10-06 07:25
刘博 2026/10/06 07:25

邦小白快读

EN
全文速览

雷军重金招揽的95后女生罗福莉,用10个月时间登顶小米职级天花板,创下小米史上最快晋升纪录。她的经历和职场心得能给普通人带来实在启发。

总:从北师大本科到北大硕士,从阿里达摩院到DeepSeek再到小米,罗福莉的路径没有捷径,而是一个低起点者靠长期主义实现技术复利的过程。

1. 晋升看点:2025年11月入职小米MiMo团队,此后发布新模型、扩大管辖范围,如今晋升22级,对应副总裁级别。小米19级员工综合年薪已超200万,22级薪酬更高,说明核心技术人才回报惊人。

2. 成长路径:她自认计算机领域绝对的低起点者,大二开始为保研北大做准备;研究从词义消歧转向文本生成,在ACL发表8篇论文,毕业后参与DeepSeek-V2等模型研发,持续在正确方向积累。

3. 实操建议:罗福莉建议年轻人保持探索欲与好奇心,极致运用前沿AI工具,在持续试错中培养专属判断力与科研审美。她强调自己并非天才,靠的是好奇心和一点点不服输的韧劲。

小米通过招揽顶尖AI人才加速大模型自研,其模型发布和定价策略对品牌商有重要启示。

总:小米MiMo团队发布并开源Xiaomi MiMo-V2.6系列,覆盖Pro与Flash两个原生全模态模型,披露了模型性能、成本优势和研发方向。

1. 产品研发:MiMo-V2.6-Pro在Artificial Analysis Intelligence Index中取得46分,位列开源权重模型榜首,研发聚焦RSI递归自我改进路径,这代表品牌可借助前沿AI能力提升产品智能化水平。

2. 价格竞争:同等智能水平下,该模型价格仅为海外模型的1/20至1/60,以极低算力成本形成竞争壁垒,品牌商可以此降低AI功能嵌入终端产品的成本,从而支撑更具竞争力的定价。

3. 品牌与生态:罗福莉在小米人车家全生态合作伙伴会亮相,小爱同学组织架构调整后基础模型能力由MiMo团队提供,说明AI正成为全生态体验的核心卖点。品牌商应关注如何将全模态模型能力与自身产品、服务结合,用技术优势带动品牌心智。

AI人才争夺加剧和开源模型生态为卖家带来新的增长机会,需关注模型能力落地和商业模式变化。

总:小米开源MiMo-V2.6系列,Pro和Flash均为原生全模态模型,且Pro位列开源权重模型榜首,这会降低AI应用开发门槛,催生新生意。

1. 增长市场:开源模型让卖家不必自研大模型即可获得顶尖能力,可围绕语音助手、智能客服、内容生成等方向开发应用,尤其全模态模型支持更多交互场景,市场需求广阔。

2. 消费需求变化:大模型正从Chat向Agent范式转变,小爱同学等智能助手将具备更强基础模型能力,消费者对智能终端的体验要求会快速提高,卖家需提前布局AI驱动的产品与服务。

3. 风险与机会:AI人才薪资连年翻倍,头部企业战略放权、职级破格,显示人才密度决定AI企业核心竞争力。卖家若依赖外部AI能力,需关注模型供应商的持续投入和价格变化;也可以借助开源模型和低价API,以轻资产方式切入垂直场景,用小团队做出差异化应用。

文章呈现AI大模型开源和全模态技术趋势,工厂可从中寻找产品智能化与数字化升级的切入点。

总:小米发布全模态模型Xiaomi MiMo-V2.6系列,支持Pro与Flash两个版本,性能领先且价格低廉,为工厂产品加AI提供现实路径。

1. 产品生产与设计需求:全模态模型意味着语音、图像等多模态交互能力更容易集成到硬件中,工厂可考虑在智能家居、车载设备、工业终端中嵌入此类模型,提升产品的交互体验和市场卖点。

2. 商业机会:MiMo-V2.6-Pro价格仅为海外模型的1/20至1/60,开源权重模型降低了AI使用门槛。工厂可基于低成本模型开发差异化智能产品,避免被单一供应商锁定,同时通过模型迭代快速更新产品功能。

3. 数字化启示:罗福莉从自然语言处理转到大模型,强调团队协作和工程能力,说明数字化和智能化升级需要复合型人才梯队与持续技术投入。工厂应关注模型代际变化带来的连接与决策能力增强,把AI视为生产端和产品端的共同基础设施。

AI大模型技术快速迭代、人才重仓模式开启,服务商需抓住模型开源化和低成本化带来的解决方案机会。

总:小米MiMo-V2.6系列展示的全模态、RL算力扩展和RSI路径,代表行业技术新趋势,而低价开源模型直击客户AI落地成本痛点。

1. 行业趋势:模型从语言走向物理世界,Agent范式正在取代传统Chat范式,客户对智能决策和自动执行的需求会爆发。服务商应紧跟模型能力边界,提前储备全模态应用方案。

2. 新技术:MiMo-V2.6-Pro在AA综合智能指数中位列开源权重模型榜首,说明开源模型性能已经可比肩海外顶尖,而Flash版本提供更轻量选择。服务商可基于这些模型快速搭建行业解决方案。

3. 客户痛点和解决方案:同等智能水平下价格仅为海外模型1/20至1/60,直接解决中小企业用不起、不敢用大模型的痛点。同时,客户普遍缺乏AI人才,服务商可提供模型微调、私有化部署、业务集成和运维服务,帮助客户在Agent范式下重构流程,形成持续服务收入。

小米的组织调整和生态大会显示平台正把大模型作为基础能力,平台商可借鉴AI人才激励与生态合作打法。

总:小米以千万年薪和职级破格绑定顶尖AI人才,并通过人车家全生态建设推动模型能力输出,这对平台吸引人才、搭建生态有参考价值。

1. 平台做法:小米将小爱同学技术能力拆分为基础模型、云端和端侧三部分,基础模型能力由MiMo团队提供,说明平台要将AI视为核心基础设施,成立专职团队并快速放权。

2. 招商与管理:罗福莉用10个月升至22级,创下最快晋升纪录,表明平台要用极致激励绑定核心技术人才,同时开放开源模型可以吸引开发者和合作伙伴,形成生态协同,扩大平台影响力。

3. 风险规避:大模型技术迭代极快,平台不能单点依赖某一个外部模型,应像小米一样自研+开源并行,关注人才流动和模型合规风险,保持研发节奏,避免因核心人员变动或技术路线偏差影响平台稳定性。

罗福莉晋升案例折射出中国AI产业人才竞争和青年技术人才地位上升的结构性变化,值得深入研究。

总:AI人才争夺已从高薪抢人升级为战略放权、资源倾斜和职级破格的全方位人才重仓模式,青年技术人才成为决定AI企业核心竞争力的关键变量。

1. 产业新动向:头部企业争抢顶尖博士和青年研究员,腾讯任命27岁前OpenAI研究员为首席AI科学家,字节启动Top Seed校招计划,百度任命1997年出生的博士为基础模型研发部负责人,小米10个月将31岁女将提到副总裁级,显示人才密度正取代单纯算力竞赛。

2. 新问题:顶尖人才快速晋升带来组织管理挑战,可能引发资历与能力的平衡、内部激励失衡、人才泡沫等问题;同时,天才少女标签与罗福莉去神化自我认知形成张力,提示成功归因的复杂性。

3. 政策与启示:研究应关注青年科技人才评价体系,弱化天才叙事,强调长期积累、工程能力和团队协作。罗福莉的路径表明,兴趣驱动、方向判断和持续试错是人才成长的关键,可为高校培养和企业用人政策提供参考。

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

Luo Fuli, the post-95s researcher Xiaomi CEO Lei Jun hired with a blockbuster package, has reached the top of Xiaomi's grade system in just 10 months—the fastest promotion in the company's history. Her experience and workplace insights offer practical takeaways for ordinary professionals.

Overview: From an undergraduate degree at Beijing Normal University to a master's at Peking University, from Alibaba's DAMO Academy to DeepSeek and then Xiaomi, Luo's path has taken no shortcuts. It is a story of a low-starter accruing technical compound returns through long-termism.

1. What stands out about the promotion: Luo joined Xiaomi's MiMo team in November 2025; since then she has released a new model, expanded her remit, and has now been elevated to Grade 22, equivalent to vice president level. At Xiaomi, Grade 19 employees already earn over RMB 2 million in total annual compensation, and Grade 22 pays considerably more—evidence that core technical talent commands extraordinary returns.

2. The growth path: Luo describes herself as an absolute low-starter in computer science. She began preparing for admission to Peking University's graduate program in her sophomore year, shifted her research from word-sense disambiguation to text generation, and published eight papers at ACL. After graduating, she worked on models including DeepSeek-V2, steadily accumulating in the right direction.

3. Practical advice: Luo advises young people to stay curious and thirsty for exploration, push frontier AI tools to their limits, and develop their own judgment and research taste through persistent trial and error. She insists she is no genius—what got her here were curiosity and a bit of stubborn tenacity.

By hiring top AI talent to accelerate in-house large-model development, Xiaomi's model release and pricing strategy carry important lessons for brand owners.

Overview: Xiaomi's MiMo team released and open-sourced the Xiaomi MiMo-V2.6 series, featuring two natively full-modal models, Pro and Flash, and disclosing details on model performance, cost advantages, and R&D direction.

1. Product development: MiMo-V2.6-Pro scored 46 on the Artificial Analysis Intelligence Index, ranking first among open-weight models. Its R&D centers on the RSI (recursive self-improvement) path, signaling that brands can harness frontier AI capabilities to make their products significantly smarter.

2. Price competition: At the same intelligence level, this model is priced at just 1/20 to 1/60 of overseas models, creating a competitive moat through extremely low compute costs. Brands can leverage this to cut the cost of embedding AI into end products, supporting more competitive pricing.

3. Brand and ecosystem: Luo Fuli appeared at Xiaomi's Human-Vehicle-Home ecosystem partner conference, and after Xiao Ai's organizational restructuring, the MiMo team now supplies its foundational model capabilities. This makes clear that AI is becoming the core selling point of the entire ecosystem experience. Brands should focus on how to combine full-modal model capabilities with their own products and services, translating technical advantage into stronger brand perception.

The intensifying competition for AI talent and the open-source model ecosystem are creating new growth opportunities for sellers, who should watch how model capabilities are being commercialized and how business models are changing.

Overview: Xiaomi has open-sourced the MiMo-V2.6 series; both Pro and Flash are natively full-modal, and Pro ranks first among open-weight models. This lowers the threshold for AI application development and is likely to spawn new businesses.

1. Growth market: Open-source models give sellers access to frontier capabilities without building their own large models. Sellers can develop applications in areas such as voice assistants, intelligent customer service, and content generation—full-modal models in particular support far more interactive scenarios, promising broad market demand.

2. Changing consumer expectations: Large models are shifting from the Chat paradigm to the Agent paradigm. Smart assistants such as Xiao Ai will have substantially stronger foundational capabilities, and consumer expectations for intelligent devices will rise quickly. Sellers need to position AI-driven products and services ahead of the curve.

3. Risks and opportunities: AI talent salaries are doubling year after year, while leading companies are delegating strategic authority and breaking grade conventions—a sign that talent density is what determines an AI firm's competitiveness. Sellers that depend on external AI capabilities should monitor model suppliers' ongoing investment and pricing changes. They can also tap open-source models and low-cost APIs to enter vertical scenarios on an asset-light basis, using small teams to build differentiated applications.

This article highlights the dual trends of open-source AI and full-modal technology. Factories can use these developments as entry points for product intelligence and digital upgrades.

Overview: Xiaomi has released the full-modal Xiaomi MiMo-V2.6 series, available in Pro and Flash versions. With leading performance and low pricing, it offers a realistic path for adding AI to factory products.

1. Product production and design needs: Full-modal models make voice, image, and other multimodal interactions far easier to integrate into hardware. Factories can consider embedding these models in smart-home devices, automotive equipment, and industrial terminals, improving the interactive experience and boosting market appeal.

2. Business opportunities: MiMo-V2.6-Pro costs only 1/20 to 1/60 as much as overseas models, and open-weight models significantly lower the barrier to AI adoption. Factories can build differentiated intelligent products on low-cost models, avoid lock-in to any single supplier, and rapidly refresh product functionality as models iterate.

3. Digitalization insights: Luo Fuli's transition from natural language processing to large models—and her emphasis on team collaboration and engineering capability—shows that digital and intelligent upgrades require a composite talent pipeline and sustained technical investment. Factories should watch how generational model changes strengthen connectivity and decision-making, and treat AI as common infrastructure on both the production side and the product side.

As large-model technology iterates rapidly and heavyweight talent strategies take hold, service providers should seize the solution opportunities created by open-source releases and falling costs.

Overview: The Xiaomi MiMo-V2.6 series showcases full-modal capabilities, RL compute scaling, and the RSI path, representing new industry trends; the low-cost open-source model directly addresses the biggest client pain point—the cost of deploying AI.

1. Industry trends: Models are moving from language to the physical world, and the Agent paradigm is replacing the traditional Chat paradigm. Client demand for intelligent decision-making and automated execution is set to surge. Service providers should track the evolving boundary of model capabilities and prepare full-modal application offerings in advance.

2. New technology: MiMo-V2.6-Pro ranks first among open-weight models on the AA (Artificial Analysis) comprehensive intelligence index, proving that open-source model performance can now match the best overseas models, while the Flash version offers a more lightweight alternative. Service providers can use these models to quickly build industry solutions.

3. Client pain points and solutions: At equal intelligence levels, the price is just 1/20 to 1/60 of overseas models, directly solving the problem of small and mid-sized enterprises that find large models too expensive and too risky to adopt. At the same time, most clients lack AI talent; service providers can offer model fine-tuning, private deployment, business integration, and ongoing operations, helping clients re-engineer processes around the Agent paradigm and creating recurring service revenue.

Xiaomi's organizational shake-up and ecosystem conference show that platforms are now making large models a foundational capability. Platform operators can borrow from its approach to AI talent incentives and ecosystem collaboration.

Overview: Xiaomi has bound top AI talent to the company with eight-figure annual salaries and exceptional grade promotions, and is pushing model capabilities outward through its Human-Vehicle-Home ecosystem. This offers a valuable playbook for platforms seeking to attract talent and build ecosystems.

1. Platform approach: Xiaomi split Xiao Ai's technical capabilities into three layers—foundational model, cloud, and on-device—with the MiMo team providing the foundational model. This signals that platforms should treat AI as core infrastructure, establish dedicated teams, and delegate authority quickly.

2. Talent attraction and management: Luo Fuli rose to Grade 22 in 10 months, a record for the fastest promotion at Xiaomi. This shows that platforms should use extreme incentives to bind core technical talent; meanwhile, open-sourcing models can attract developers and partners, generating ecosystem synergies and expanding platform influence.

3. Risk mitigation: Large-model technology evolves extremely fast, and platforms cannot depend on a single external model. They should follow Xiaomi's example of pursuing in-house development and open source in parallel, while closely monitoring talent mobility and model compliance risks. Maintaining a consistent R&D cadence is essential to avoid instability caused by changes in key personnel or deviation in technical direction.

Luo Fuli's rapid rise reflects a structural shift in China's AI industry—intensified competition for talent and the elevated status of young technical professionals—that deserves in-depth study.

Overview: The AI talent war has escalated from high-salary poaching into an all-out mode of talent-heavy investment characterized by strategic delegation, resource tilting, and grade-breaking promotions. Young technical talent has become a decisive variable in the core competitiveness of AI enterprises.

1. New industry dynamics: Leading companies are scrambling for top PhDs and young researchers. Tencent appointed a 27-year-old former OpenAI researcher as its chief AI scientist; ByteDance launched the Top Seed campus recruitment program; Baidu named a PhD born in 1997 to lead its foundation-model R&D division; and Xiaomi elevated a 31-year-old woman to vice president level in 10 months. Talent density, not sheer computing power, is now the differentiator.

2. New questions: The rapid promotion of exceptional talent creates organizational management challenges—tensions between seniority and ability, distortions in internal incentives, and the risk of a talent bubble. In parallel, the "genius girl" label and Luo Fuli's own efforts to de-mythologize herself highlight the complexity of attributing success.

3. Policy and implications: Research should focus on evaluation systems for young scientific and technical talent, downplaying genius narratives and emphasizing long-term accumulation, engineering ability, and team collaboration. Luo Fuli's trajectory suggests that interest-driven motivation, direction sense, and sustained trial and error are critical to talent development, offering useful lessons for university education and corporate hiring policies.

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.

这也许是目前雷军最器重的人。

近日小米对内发布晋升名单,31岁的大模型团队负责人罗福莉晋升至22级。这意味着从2025年11月官宣入职算起,到登顶小米职级天花板,她只用了短短10个月。

如此,创下小米史上最快晋升纪录。

这位来自四川宜宾的女生,本科毕业于北京师范大学,硕士则是北京大学,为人熟知的经历是在DeepSeek任职,曾被外界冠以“天才少女”,最终被雷军重金招揽。但她也不止一次对外表达过,自己并非天才,只是想安安静静做事情。

将视野放大来看,罗福莉的破格晋升,恰是中国AI产业人才竞争的真实缩影。这群年轻人,正在引领着中国AI产业潮水的方向。

雷军挖来95后女生,最快晋升速度

10个月,罗福莉的小米生涯势如破竹。

伴随着此次晋升消息,还有MiMo团队正式发布并开源Xiaomi MiMo-V2.6系列,包含Pro与 Flash两个原生全模态模型。得益于RL算力的扩展,MiMo-V2.6-Pro在 Artificial Analysis Intelligence Index(AA综合智能指数)中取得46分,位列开源权重模型榜首;且在同等智能水平下,价格仅为海外模型的1/20至 1/60。

这被MiMo团队视为探索RSI(递归自我改进)路径的关键一步。罗福莉坦言,“这次工作背后的研究创新和工程挑战,超过了我曾部分参与的DeepSeek R1。”

据公开信息显示,除去小米创始人雷军之外,小米集团内部共设13到22十层职级,其中总监对应19至20级,副总裁对应21至22级,且小米19级员工的综合年薪能够达200万元以上,这意味着22级的薪酬势必更高。换言之,罗福莉在带队完成新模型发布后,已跻身小米核心高管行列。

追溯罗福莉加入小米的始末,还要从一则热搜说起。

去年初,据《证券时报》报道,雷军曾希望用千万年薪挖角罗福莉,后者由此意外走红。随后在去年11月,她在朋友圈发文“智能终将从语言迈向物理世界”,宣布自己加入小米MiMo团队。

很快,罗福莉就完成在小米的公开首秀。在去年12月的小米“人车家全生态”合作伙伴会上,她以MiMo大模型负责人的身份正式亮相。

今年3月,小米一口气发布三款模型,包括面向agent时代的旗舰基座模型Xiaomi MiMo-V2-Pro,以及全模态基座模型Xiaomi MiMo-V2-Omni和语音合成模型Xiaomi MiMo-V2-TTS。

彼时,罗福莉在社交媒体发文,“我称这是悄无声息的伏击——不是因为我们有计划,而是因为从Chat到Agent范式的转变发生得如此之快,甚至我们自己都几乎不敢相信。中间某个地方,有一个过程既令人兴奋、痛苦,又迷人。”

今年7月,小米调整小爱同学组织架构,将技术能力拆分为基础模型、云端和端侧三部分。其中,基础模型能力由罗福莉负责的MiMo团队提供,这也表明她的业务管辖范围进一步扩大。直至此次晋升,罗福莉用短短10个月时间,登顶了小米职级天花板。

雷军对年轻AI人才的渴求,不言而喻。

梁文锋留不住她,罗福莉成长史

罗福莉身后,并非是外界想象中的那种年少成名的传奇故事。

1995年出生在宜宾的一个普通家庭,罗福莉高中就读于宜宾市第一中学,她的班主任曾评价“不算拔尖,但足够刻苦。”后来她考入北京师范大学计算机专业,用她自己的话来说是“计算机领域绝对的‘低起点者’”。即便如此,罗福莉在大二时就开始为保研北京大学计算机专业做准备,经过努力最终如愿。

起初,她的研究兴趣聚焦于自然语言处理(NLP)中一个细分领域——词义消歧。但随着大模型技术的兴起,她判断传统消歧任务的本质,正被大模型的“下一个词预测”范式所覆盖。于是,她果断将研究重心转向文本生成,这也为她日后研发大模型埋下了伏笔。

罗福莉真正崭露头角的一年是2019年,当时她在自然语言处理领域的顶级国际会议ACL上发表了8篇论文,其中2篇为第一作者。“天才少女”的标签也随之而来。

但对这一标签,罗福莉却有着近乎本能的排斥。她曾在北大校友会的访谈中直言,这类标签是为了吸引注意力而创造的刻板印象,“它粉饰了科研过程中真正的努力、坚韧和全心投入,暗示成功是天生的。”

回忆起在北大读研的这几年,罗福莉曾感慨,“北大自由的科研环境让我受益匪浅。”同时,北大带给她的,还有最早的“协作启蒙”。在北大的实验室里,她初次体会到了Teamwork(团队协作)的力量:有同学擅长搭建GPU集群,有人专攻算法,大家各展所长。

硕士毕业后,罗福莉职业生涯的第一站选在了阿里。她通过“阿里星”入职阿里达摩院,担任机器智能实验室研究员,从事人工智能领域研究。在这里,她主导开发了多语言预训练模型VECO,并推动了AliceMind的开源工作。

三年后,罗福莉加入DeepSeek母公司幻方量化从事深度学习相关工作,后又担任DeepSeek的深度学习研究员,参与研发DeepSeek-V2等模型。伴随着DeepSeek大模型的横空出世,罗福莉开始更多地出现在聚光灯之下,她曾总结,“在DeepSeek的那段经历,是一次重大的转折。”

回过头看,罗福莉成长之路并无任何捷径、侥幸,而是一个典型的长期主义技术复利成果,且她一直坚持着去神化的自我认知。正如她所言,自己并非是天才,起点其实很普通,“是对本质问题探索的好奇心、不错的行动力以及一点点不服输的韧劲,才一步步走到今天。”

AI大战,高材生黄金时代

一场AI人才争夺战悄然浮现。

不禁让人想起此前一幕——去年12月,腾讯任命前OpenAI研究员姚顺雨为首席AI科学家,此时后者年仅27岁;随后混元多模态模型部门与大语言模型部门合并,成立基础模型部,统一由这位年轻人管理。

字节跳动同样火力全开。今年初,DeepSeek核心研究员郭达雅离职,引发一众互联网大厂的追逐,最终他选择加入字节Seed部门。与此同时,字节跳动启动“Top Seed”校招计划,面向全球招募30名顶尖应届博士,强调提供行业Top级薪酬、不设上限。

百度也加入到这场争夺——今年7月,1997年出生的孙天祥正式加入百度,出任基础模型研发部负责人,同时进入百度模型委员会。这位复旦大学计算机博士,是类ChatGPT开源大模型MOSS的第一核心开发者,并在国际上率先提出Model-as-a-Service(MaaS)概念。

当年轻人们还没出校园,大厂就已争先恐后递出了橄榄枝。

腾讯在今年3月正式启动“青云计划”实习生招聘,开放大语言模型、多模态、智能体等六大技术领域,其中薪酬政策上不封顶;月之暗面宣布推出Kimi顶 尖人才“穿越计划”,将授予尚未毕业的实习生公司期权;MiniMax也启动Top Talent人才计划,CEO闫俊杰更是公开放话:对真正顶尖的人才,薪资和期权上不封顶,不看资历,只看你能解决多大问题。

早在2025年末,Business Insider就披露一组数据——一份4-6个月的AI实习和短期研究项目,月薪已经来到7000-18000美元区间,折合人民币约4.9-12.6万元。据晚点LatePost报道,2024年字节Top Seed校招生年薪约150万元,2025年涨至300万-500万元,2026年部分核心岗位更是开到了600万元,连年翻倍。

毫无疑问,国内的AI人才争夺已完成全面升级,从早期的高薪抢人、福利内卷,进阶为战略放权、资源倾斜、职级破格的全方位人才重仓模式。这也标志着,青年技术人才已然成为决定AI企业核心竞争力,影响产业格局走向。

而罗福莉、姚顺雨等人的路径给年轻人提供的参考,恰恰不是“天才叙事”,而是一条更朴素的逻辑:在正确的方向上持续积累工程能力,把前沿技术做到可落地,然后在一个足够大的平台上放大自己的价值。

正如罗福莉在智源大会上给年轻人的建议:始终保持探索欲与好奇心。“极致运用前沿AI工具,在持续试错的过程中,培养专属自己的判断力与科研审美。在技术高速变革的时代,独特的认知、判断与审美,是年轻人最核心、最不可替代的竞争力。”

归根结底,AI的竞争从来不只是模型与算力的竞赛,而是人才密度的比拼。

可以笃定的是,下一代AI版图的更改,不会是在某一家企业的发布会上,而是会发生在这些年轻人决定“做什么、不做什么”的那些时刻。他们选择的方向,就是中国AI的方向。

注:文/刘博,文章来源:投资界(公众号ID:pedaily2012),本文为作者独立观点,不代表亿邦动力立场。

文章来源:投资界

广告
微信
朋友圈

FAQ回顾

罗福莉是谁?

罗福莉是小米大模型团队负责人,1995年生于四川宜宾,本科毕业于北京师范大学,硕士毕业于北京大学。她曾任阿里达摩院研究员、DeepSeek深度学习研究员,参与研发DeepSeek-V2等模型。2025年11月加入小米MiMo团队,2026年晋升22级,成为小米核心高管。

罗福莉为什么能在10个月内晋升到小米22级?

罗福莉在AI大模型领域积累深厚:曾在ACL发表8篇论文,主导开发VECO,参与DeepSeek-V2研发。加入小米后,她带队发布Xiaomi MiMo-V2系列和MiMo-V2.6系列全模态模型,MiMo-V2.6-Pro在AA综合智能指数中获46分、位列开源权重模型榜首,并因带队完成新模型发布而晋升至22级。

小米MiMo-V2.6是什么,有什么特点?

小米MiMo-V2.6是MiMo团队发布并开源的全模态模型系列,包含Pro与Flash两个版本。MiMo-V2.6-Pro在Artificial Analysis Intelligence Index(AA综合智能指数)中取得46分,位列开源权重模型榜首;同等智能水平下,价格仅为海外模型的1/20至1/60,被团队视为探索RSI(递归自我改进)路径的关键一步。

国内AI大模型顶尖人才的薪资有多高?

AI人才薪资涨幅惊人:Business Insider披露,2025年末4-6个月AI实习月薪为7000-18000美元;晚点LatePost报道,字节Top Seed校招年薪2024年约150万元,2025年涨至300万-500万元,2026年部分核心岗位达600万元。腾讯、月之暗面、MiniMax等也推出薪酬或期权不设上限的人才计划。

为什么AI企业都在争抢顶尖年轻人才?

因为AI竞争不只是模型与算力的竞赛,更是人才密度的比拼。腾讯任命27岁姚顺雨为首席AI科学家,字节吸纳DeepSeek核心研究员郭达雅,百度聘请1997年出生的孙天祥负责基础模型部。青年技术人才已被视为决定AI企业核心竞争力的关键,其选择将影响中国AI产业格局走向。

这么好看,分享一下?

朋友圈 分享

APP内打开

赞 +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