#GoogleAIReshuffle

99,3 tys. oglądających|90 wpis

About GoogleAIReshuffle

Alphabet enacted a major shake-up in its core AI ranks: DeepMind founder Demis Hassabis has stepped down from day-to-day management, Google AI veteran Jeff Dean announced his departure to found Discovery Loop, and several senior executives tied to Gemini have also exited. This marks a pivotal leadership reset that will likely reshape Google's AI roadmap.

GoogleAIReshuffle Popularne wpisy

Przypięte
Blockbeats
Blockbeats
BadaniaHassabis卸任,Jeff Dean离职创业,谷歌的军心散了
原文标题:《Hassabis卸任,Jeff Dean离职创业,谷歌的军心散了》 原文作者:动察Beating 8 月 5 日,Google DeepMind CEO Demis Hassabis 卸下日常管理职务,转任 DeepMind 董事长,并出任 Alphabet 首席科学家。他仍然负责长期 AI 研究,也继续领导新药研发公司 Isomorphic Labs。头衔比过去更高,但其实是一次明升暗降,手里的实际管理权变少了。 Gemini 的研发和交付将由原 DeepMind 首席技术官 Koray Kavukcuoglu 负责,直接向皮查伊汇报。Hassabis 仍然是谷歌 AI 的科学象征,继续参与 AGI 方向的判断,却不再掌握 DeepMind 的日常运转。经过这次调整,DeepMind 与 Alphabet 总部之间的管理链条进一步缩短,皮查伊也将更直接地掌握 Gemini 的研发进度。 同一天,Jeff Dean 宣布离开谷歌。 他带走了 Sanjay Ghemawat、Oriol Vinyals 和 Quoc Le,成立一家名为 Discovery Loop 的公司,希望用 AI 自动完成提出问题、设计实验、执行实验和评估结果的整个科研流程。谷歌参与了创始投资,将成为它的云服务商,并提供第一年的算力。 谷歌没有强留这支团队,而是选择继续保持资本、云服务和科研合作上的关系。对于一批已经决定创业的顶尖研究者来说,这算得上体面的处理方式。 市场仍然迅速做出了反应。Alphabet 股价盘中一度下跌超过 5%,收盘跌幅接近 4%。以这家公司的体量计算,3.8% 的跌幅对应约 1750 亿美元的市值,盘中下跌最多时,蒸发规模接近 2600 亿美元。 过去六个星期,这已经是 Alphabet 第四次因为 AI 受到市场惩罚。 Jeff Dean 与 Hassabis,谷歌 AI 的两种传统 Jeff Dean 很难被归入一般意义上的技术高管。 1999 年加入谷歌时,他是公司第 30 号员工。此后近三十年,谷歌几次最重要的技术转型,他几乎都在场。 早期搜索和广告系统需要处理迅速膨胀的数据,他和 Ghemawat 写出了 MapReduce 和 Bigtable。这些系统后来成为大规模分布式计算的基础,影响远远超出谷歌。Google Brain、TensorFlow 和 TPU 的建立,同样有他的参与。到了 Gemini 时代,他又成为模型的技术联合负责人。 谷歌今天能够训练大模型,能够在自研芯片上运行模型,也能够把模型接入搜索、云和各种消费产品,背后都能找到他参与修建的基础设施。 Hassabis 代表的是另一种能力。 2010 年,他在伦敦创办 DeepMind。2014 年,谷歌完成收购。两年后,AlphaGo 击败李世石,人工智能第一次以如此直接的方式进入全球公众视野。2024 年,他与 John Jumper 凭借 AlphaFold 获得诺贝尔化学奖。 2023 年,Google Brain 与 DeepMind 合并,Hassabis 接管整合后的 Google DeepMind。他既是 Gemini 项目的领导者,也是谷歌对外讲述 AI 未来时最重要的人物。 Dean 和 Hassabis,分别支撑着谷歌 AI 的工程传统与科学传统。如今,Dean 离开公司,Hassabis 退居二线。 六个星期里的四次暴跌 这次调整之所以引起如此强烈的市场反应,还因为它发生在谷歌 AI 连续受挫之后。 6 月,Transformer 论文作者之一、Gemini 技术联合负责人 Noam Shazeer 离开谷歌,加入 OpenAI。2024 年,谷歌刚刚通过一笔约 27 亿美元的特殊交易,将他从 Character.AI 请回公司,并安排他进入 Gemini 核心团队。两年后,Shazeer 再次离开。 随后,AlphaFold 负责人、2024 年诺贝尔化学奖得主 John Jumper 加入 Anthropic。 两天内,谷歌失去了两位极有分量的研究者。6 月 22 日,Alphabet 股价盘中一度下跌超过 7%,市值蒸发超过 2000 亿美元。 7 月 16 日,Gemini 3.5 Pro 延期的消息传出。这款旗舰模型原计划在 6 月发布,因为代码能力等指标未达到预期,进度已经推迟数月。消息公布后,Alphabet 股价盘中再次下跌超过 3%。 7 月下旬,压力转移到了财务报表上。 Alphabet 第二季度资本开支达到 449 亿美元,同比翻倍,全年资本开支指引被上调到 1950 亿至 2050 亿美元。同期,公司自由现金流转为负 59 亿美元,这是 Alphabet 历史上第一次出现季度自由现金流为负。财报发布后的第一个交易日,股价盘中跌幅超过 7%。 谷歌投入的资源越来越多,模型却没有形成与投入相称的领先优势。而且旗舰产品没有按时交付出来,负责研究和工程的核心人物又在相继离开。 Jumper 离开后,有媒体援引 DeepMind 员工的说法称,文本、图像、视频、语音和视觉等领域,谷歌已经很难找到明确处于行业最前沿的模型。 谷歌从不缺资源,甚至可以说,它拥有整个 AI 行业最完整的一套资源。 它有自研 TPU,有全球规模的数据中心,有搜索和 YouTube 积累的数据,有 Android、Chrome、Workspace 和云服务提供的分发渠道。皮查伊也在内部信中列出了一串成绩,说 Gemini 应用月活用户达到 9.5 亿,Gemma 下载量超过 9 亿,Gemini Robotics 也在继续推进。 这些资产足以证明谷歌仍然强大,却无法代替下一代旗舰模型本身。市场愿意为长期投入付钱,前提是这笔钱最终能够换来领先。模型延期与核心人员离开连续发生后,投资人开始怀疑,谷歌缺少的究竟是时间,还是把庞大资源拧成一股绳的能力。 DeepMind 告别创始人管理 接替 Hassabis 负责日常运营的 Koray Kavukcuoglu 在 DeepMind 工作了 13 年,师从杨立昆,参与过 DQN、WaveNet 和多代 Gemini 项目,过去担任 Google DeepMind 的 CTO。 Koray 熟悉 DeepMind,也长期负责连接研究、基础设施和产品。由他接管 Gemini,说明谷歌接下来更重视模型交付、产品协同和商业落地。 这次调整还改变了 DeepMind 在谷歌内部的位置。 Hassabis 时代的 DeepMind 始终保留着鲜明的创始人色彩。它从伦敦起家,被谷歌收购后仍保持相对独立的研究传统。Google Brain 与 DeepMind 合并之后,Hassabis 成为整合团队的 CEO,也让 DeepMind 的文化在新组织中占据主导地位。 如今,Hassabis 退出日常管理,Koray 直接向皮查伊汇报,DeepMind 与 Alphabet 总部的关系变得更紧密。对谷歌来说,这是一次加强控制的调整。皮查伊需要更清楚地掌握 Gemini 的研发节奏,也需要有人对模型延期、资源投入和产品落地承担直接责任。 对 DeepMind 而言,这也意味着一个阶段的结束。 Hassabis 仍然在谷歌,仍然拥有崇高的头衔,也仍然参与长期研究。但那家由创始人直接领导、以科学理想塑造组织性格的 DeepMind,正在逐渐成为 Alphabet 内部更加标准化的一支 AI 研发部门。 谷歌正在学习如何送走人才 Jeff Dean 的离职,展示了谷歌处理人才外流的另一种方式。 谷歌参与 Discovery Loop 的创始投资,继续提供云服务和算力,并与新公司保持研究合作。Dean 和他的团队虽然离开了组织,仍然处在谷歌的资本和技术生态之内。 这项安排对双方都有好处。 Discovery Loop 获得了昂贵的算力、基础设施和早期资金,不必从零开始搭建研究平台。谷歌则保留了投资回报、云订单和未来合作的可能,也避免把一批工作二十多年的核心人才直接推向竞争对手。 从公司治理角度看,这几乎已经是理想的离职方案。 但它仍然无法回答一个问题,这样一支熟悉谷歌、参与建造谷歌、也能够调用谷歌资源的团队,为什么最终认为,离开以后更容易完成自己的研究? 他们没有放弃过去的研究方向,也没有因为内部竞争失败而被迫出走。他们只是将原本在谷歌内部进行的工作,搬到一家自己能够完全掌控的新公司。 他们在谈到离开的原因时说,在大组织里,总有太多惯性需要克服,才能推动激进的变化。他们想造一些不同的东西。 谷歌过去吸引顶尖研究者,靠的是人才留在这里可以获得足够多的资源,完成其他地方无法完成的工作。 如今,一批最了解谷歌资源的人开始做出相反判断。他们依然相信那些研究值得投入,也依然相信 AI 将改变科学,只是不再相信大公司的组织方式最适合完成它。 实验室之外,新的公司开始生长 谷歌 DeepMind 让人想起贝尔实验室。 贝尔实验室依靠 AT&T 长期稳定的利润,汇集了二十世纪最优秀的一批科学家和工程师,先后诞生晶体管、激光、信息论和 Unix。它的伟大没有随着组织本身完整保存下来,却通过离开的人不断扩散。 晶体管联合发明人 William Shockley 离开贝尔实验室后,在加州创办肖克利半导体。1957 年,八名工程师集体离开,成立仙童半导体。Intel 和 AMD 的创始团队,此后又从仙童体系中分化出来。后来被称为硅谷的产业网络,很大一部分正是沿着这条人才流动的路径形成。 谷歌 DeepMind 还没有走到贝尔实验室的结局。Gemini 拥有庞大的用户,Google Cloud 仍在增长,两者的相似之处在于,一家依靠成熟商业体系供养的顶级实验室自己培养出来的人,开始到外面建立新的实验室和公司。 一家实验室最辉煌的时候,人们看它聚集了多少天才。走到下一个阶段,人们会看这些天才离开后又建成了什么。 这是一种成就,也是一种损失。对整个科技行业而言,两者可以同时成立,但是对谷歌而言,这笔账很难如此轻松地一笔勾销。 谷歌投资 Discovery Loop,可以分享这家公司未来的收益,为它提供云服务也能把算力支出重新变成收入。可一家公司能够投资外部成果,却无法借此替代内部失去的创造力。 Dean 在谷歌工作了 27 年。他说,自己大约五周前才开始认真考虑创业。 一个在公司待了近三十年的人,用五周决定离开。Discovery Loop 里仍然是那些熟悉的搭档,Ghemawat 负责系统,Vinyals 负责研究,Quoc Le 继续做自动化机器学习。人员分工没有发生太大变化,研究方向也没有突然改变,只是他们从谷歌的大楼里搬了出去。 Hassabis 在告别信中说,AGI 已经近在眼前。Dean 创办的新公司,同样押注 AI 将改变科学研究。 他们大概仍然相信同一个未来,只是开始各自奔前程。 原文链接
TBNG_OKX
TBNG_OKX
#GoogleAIReshuffle Google's AI Leadership Reset Could Shape the Next Phase of the AI Race Alphabet has announced one of its biggest AI leadership shake-ups since generative AI entered the mainstream. DeepMind founder Demis Hassabis is stepping back from day-to-day management, while longtime Google AI leader Jeff Dean is leaving to launch a new venture. Several senior executives associated with Gemini have also departed. Leadership changes of this scale rarely happen in isolation. They often signal a broader shift in strategy. Google remains one of the world's most important AI companies, but competition has intensified dramatically over the past two years. OpenAI, Anthropic, Meta and xAI are all investing aggressively in talent, infrastructure and frontier models. The next phase of the AI race may be less about releasing the next chatbot. It could be about execution. How organizations allocate capital, retain top researchers and translate breakthroughs into products may become the defining competitive advantage. Technology leadership has always been cyclical. The companies that adapt fastest tend to define the next era. Google's latest reshuffle may not be the end of one chapter. It could be the beginning of another. Do you think leadership changes will strengthen Google's AI strategy, or create more uncertainty during an already competitive period? Share your thoughts below 👇
Felix.Crypto
Felix.Crypto
Google AI Reshuffle: A New Catalyst for the AI Cycle? Alphabet has announced its biggest AI leadership reshuffle in years, signaling that the next phase of the AI race will be defined by execution speed and commercialization. Demis Hassabis will shift his focus toward long-term AGI strategy and frontier AI research, while the new leadership team takes responsibility for accelerating Gemini's product rollout. At the same time, AI pioneer Jeff Dean and several senior researchers have left Google to launch a new AI startup, highlighting the intensifying competition for top AI talent. The move comes as Google faces increasing pressure from OpenAI, Anthropic, Meta, and xAI. The battle is no longer just about building the most powerful models—it's about scaling AI infrastructure, expanding data centers, and deploying advanced computing capacity fast enough to meet exploding demand. If Google successfully accelerates its AI investments, demand for GPUs, HBM, enterprise SSDs, NAND, and next-generation memory solutions could strengthen further. These technologies remain the backbone of training and serving increasingly complex AI models. That is why investors continue to watch memory leaders such as $SNDK and $SKHYNIX. Despite recent share-price weakness, the long-term AI infrastructure story remains intact. Google has also joined the High Bandwidth Flash (HBF) ecosystem, reinforcing expectations that next-generation memory technologies will play a critical role in future AI workloads. With Big Tech expected to continue raising AI capital expenditure in the second half of the year, semiconductor and memory companies could once again become key beneficiaries of the AI investment cycle. The question now is: If Google's restructuring successfully reignites its AI momentum, will $SNDK and $SKHYNIX finally respond with a meaningful rebound, or will investors need stronger evidence that AI infrastructure spending is translating into higher earnings? #GoogleAIReshuffle #RussiaCryptoLawSep1 #SandiskBeatAndBuyback $SNDK $SKHYNIX
Renee_OKX
Renee_OKX
#GoogleAIReshuffle Alphabet is reorganizing its AI leadership. Demis Hassabis is stepping away from the daily management of Google DeepMind to become its chairman and Alphabet’s chief scientist. Jeff Dean and several prominent Google AI researchers are also leaving to establish a new company called Discovery Loop. The changes could allow Hassabis to focus more deeply on AGI and scientific research, but the simultaneous departure of senior talent introduces execution risk. Investors should watch whether Gemini development slows, whether additional researchers leave, and how effectively the new management structure coordinates research with consumer products. In AI, organizational speed can be almost as important as model quality.
TrendSpider
TrendSpider
BREAKING: Google $GOOGL stock drops 5% in minutes after longtime Google chief scientist Jeff Dean announces his departure from the company. Demis Hassabis will step down as CEO of DeepMind to assume the role as the unit's Chairman.
Nathan Lambert
Nathan Lambert
Major restructuring at Gemini (Dean out, Hassabis no longer CEO). This story will be studied forever as the incumbant with all the advantages not being able to get going. P.s. OpenAI accomplished their original goal.
FinancialJuice
FinancialJuice
Google's chief Scientist is leaving to start his own AI company - WSJ. $GOOGL
Giacomo Zucco (Bear Market Edition)
Giacomo Zucco (Bear Market Edition)
I just resigned from the board of OCEAN. I still love Luke as a person very much, and I think he's following his ideals and heart sincerely. I still think miner-side block production (be it DATUM or SV2), KYC-free coinbase-payout and LN share markets are brilliant and very needed ideas for Bitcoin mining. I have honestly tried my best to serve and help until possible, but I also need to speak the truth.🧡
bradford cross
bradford cross
imho Google is having other miss like when Ilya left in 2015. So much top talent leaving because they want to work on RSI and the org doesn’t make sense / doesn’t support the different bets. Here’s what I think … They needed to fork; 1) move Sundar back to CEO of only Google, 2) founders step back in to lead Alphabet, 3) spin deepmind back out to top level under Alphabet, and let Demis work on world models, 4) let the rest of the guys form a new AI startup under Alphabet and fund it with some billions, 5) probably leave Gemini as a dedicated parallel LLM effort within Google to integrate throughout the Google products. Not seeing this and trying to mash it all together into one under Google with Sergey diving in there instead of at alphabet level supporting a few different bets is going to prove to be a poor choice. Instead many Google people left and now this Discovery Loop team isn’t an explicit top down alphabet bet on equivalent levels, which it should be. Demis stepping back from DeepMind to focus on Alphabet-wise is a mistake — they should have bumped DeepMind back out to top level under Alphabet and has Demis outside what he thinks is best. Right now there are some different ideas about the next leaps esp world models vs RSI, and imho it’d be better to build a basket of options than go all in behind one leader. Most people (including myself) didn’t count on deep learning being the near term way forward in 2012-15 when Ilya left Google to start OpenAI, but here we are. We saw Alexander, Word2Vec, and then an avalanche. I think it’s wise to let the horses run wild and free and see what happens here. I’m not sure it’s entirely clear what the next 5-10 years look like in AI and what leaps are ahead, and there may be many parallel advances that are equally important. Just my 2c and could be wrong 🤷‍♂️
Jeff Dean
Jeff Dean

Ten post nie jest już dostępny.

YabaLeftOnline
YabaLeftOnline
JUST IN: X Head of Product Nikita Bier steps down.
non aesthetic things
non aesthetic things
Nikita Bier announces he is stepping down as Head of Product at 𝕏, will remain an advisor