#GoogleAIReshuffle

Original Title: "Hassabis Steps Down, Jeff Dean Leaves to Start a Business, Google's Morale Shattered" Original Author: Dongcha Beating On August 5, Google DeepMind CEO Demis Hassabis stepped down from daily management duties to become DeepMind's Chairman and Alphabet's Chief Scientist. He remains responsible for long-term AI research and continues to lead the drug discovery company Isomorphic Labs. Although his title is higher than before, this is actually a subtle demotion, with less actual management power in his hands. The development and delivery of Gemini will be led by former DeepMind CTO Koray Kavukcuoglu, who will report directly to Pichai. Hassabis remains the scientific symbol of Google AI and continues to participate in AGI direction decisions but no longer controls DeepMind's daily operations. After this adjustment, the management chain between DeepMind and Alphabet headquarters is further shortened, and Pichai will have more direct control over Gemini's development progress. On the same day, Jeff Dean announced his departure from Google. He took with him Sanjay Ghemawat, Oriol Vinyals, and Quoc Le to establish a company called Discovery Loop, aiming to use AI to automate the entire scientific research process of posing questions, designing experiments, executing experiments, and evaluating results. Google participated in the founding investment, will be its cloud service provider, and will provide computing power for the first year. Google did not strongly retain this team but chose to maintain capital, cloud service, and research cooperation relationships. For a group of top researchers who have decided to start a business, this is a respectable handling. The market quickly reacted. Alphabet's stock price fell more than 5% intraday and closed down nearly 4%. Given the company's scale, a 3.8% drop corresponds to about $175 billion in market value, with intraday losses peaking near $260 billion. In the past six weeks, this is the fourth time Alphabet has been punished by the market due to AI. Jeff Dean and Hassabis: Two Traditions of Google AI Jeff Dean is hard to classify as a typical technical executive. He joined Google in 1999 as the 30th employee. Over nearly thirty years, he has been present at almost every major technical transformation at Google. Early search and advertising systems needed to handle rapidly expanding data; he and Ghemawat wrote MapReduce and Bigtable. These systems later became the foundation of large-scale distributed computing, influencing far beyond Google. He also participated in the establishment of Google Brain, TensorFlow, and TPU. In the Gemini era, he became the technical co-lead of the model. Google's ability today to train large models, run models on self-developed chips, and integrate models into search, cloud, and various consumer products all trace back to infrastructure he helped build. Hassabis represents a different capability. In 2010, he founded DeepMind in London. In 2014, Google completed the acquisition. Two years later, AlphaGo defeated Lee Sedol, marking AI's first direct entry into global public awareness. In 2024, he and John Jumper won the Nobel Prize in Chemistry for AlphaFold. In 2023, Google Brain merged with DeepMind, and Hassabis took over the integrated Google DeepMind. He is both the leader of the Gemini project and the most important figure when Google talks externally about AI's future. Dean and Hassabis respectively support Google's AI engineering tradition and scientific tradition. Now, Dean has left the company, and Hassabis has stepped back. Four Sharp Drops in Six Weeks This adjustment triggered such a strong market reaction partly because it happened after a series of setbacks for Google AI. In June, Noam Shazeer, one of the Transformer paper authors and Gemini's technical co-lead, left Google to join OpenAI. In 2024, Google had just brought him back from Character.AI through a special deal worth about $2.7 billion and placed him in the Gemini core team. Two years later, Shazeer left again. Then, John Jumper, head of AlphaFold and 2024 Nobel Chemistry laureate, joined Anthropic. Within two days, Google lost two very influential researchers. On June 22, Alphabet's stock price fell more than 7% intraday, wiping out over $200 billion in market value. On July 16, news broke that Gemini 3.5 Pro was delayed. This flagship model was originally planned for release in June but was postponed by several months due to underperformance in code capabilities and other metrics. After the announcement, Alphabet's stock price fell more than 3% intraday again. In late July, pressure shifted to financial reports. Alphabet's Q2 capital expenditure reached $44.9 billion, doubling year-over-year, and the full-year capital expenditure guidance was raised to $195 billion to $205 billion. Meanwhile, the company reported a free cash flow of negative $5.9 billion for the first time in its history. On the first trading day after the earnings release, the stock price fell more than 7% intraday. Google is investing more resources, but the models have not formed a leading advantage proportional to the investment. Moreover, flagship products are not delivered on time, and core research and engineering personnel are leaving one after another. After Jumper's departure, media cited DeepMind employees saying that in fields like text, image, video, voice, and vision, Google is finding it difficult to identify models clearly at the industry's forefront. Google never lacks resources; it arguably has the most complete set of resources in the entire AI industry. It has self-developed TPUs, global-scale data centers, data accumulated from Search and YouTube, distribution channels from Android, Chrome, Workspace, and cloud services. Pichai also listed a series of achievements in an internal letter, saying Gemini has 950 million monthly active users, Gemma has over 900 million downloads, and Gemini Robotics is continuing to advance. These assets prove Google remains strong but cannot replace the next-generation flagship model itself. The market is willing to pay for long-term investment, provided that the money ultimately yields leadership. After consecutive model delays and core personnel departures, investors begin to doubt whether Google lacks time or the ability to unify its vast resources. DeepMind Bids Farewell to Founder Management Koray Kavukcuoglu, who succeeded Hassabis in daily operations, has worked at DeepMind for 13 years, studied under Li-Kun Yang, participated in DQN, WaveNet, and multiple Gemini projects, and previously served as Google DeepMind's CTO. Koray is familiar with DeepMind and has long been responsible for connecting research, infrastructure, and products. His takeover of Gemini indicates Google will focus more on model delivery, product synergy, and commercial implementation. This adjustment also changes DeepMind's position within Google. DeepMind under Hassabis always retained a strong founder's character. It started in London and maintained a relatively independent research tradition after being acquired by Google. After Google Brain merged with DeepMind, Hassabis became CEO of the integrated team, making DeepMind's culture dominant in the new organization. Now, with Hassabis stepping back from daily management and Koray reporting directly to Pichai, DeepMind's relationship with Alphabet headquarters has tightened. For Google, this is a control-strengthening adjustment. Pichai needs clearer oversight of Gemini's development pace and someone to take direct responsibility for model delays, resource investment, and product delivery. For DeepMind, this also marks the end of an era. Hassabis remains at Google with a prestigious title and continues to participate in long-term research. But the DeepMind once directly led by its founder and shaped by scientific ideals is gradually becoming a more standardized AI research department within Alphabet. Google Is Learning How to Let Talent Go Jeff Dean's departure shows another way Google handles talent outflow. Google participated in Discovery Loop's founding investment, continues to provide cloud services and computing power, and maintains research cooperation with the new company. Although Dean and his team left the organization, they remain within Google's capital and technical ecosystem. This arrangement benefits both sides. Discovery Loop gains expensive computing power, infrastructure, and early funding without building a research platform from scratch. Google retains investment returns, cloud orders, and future cooperation possibilities, avoiding directly pushing a core team with over twenty years of experience toward competitors. From a corporate governance perspective, this is almost an ideal departure plan. But it still cannot answer one question: why does such a team, familiar with Google, having helped build Google and able to access Google's resources, ultimately believe it is easier to complete their research after leaving? They have not abandoned their past research directions nor been forced out due to internal competition. They simply moved work originally done inside Google to a new company they can fully control. They said the reason for leaving is that in a large organization, there is always too much inertia to overcome to push radical change. They want to create something different. Google used to attract top researchers by offering enough resources to accomplish work impossible elsewhere. Now, a group of people who know Google's resources best are making the opposite judgment. They still believe the research is worth investing in and that AI will change science, but no longer believe a large company's organizational model is best suited to accomplish it. Outside the Lab, New Companies Begin to Grow Google DeepMind reminds people of Bell Labs. Bell Labs, supported by AT&T's long-term stable profits, gathered some of the 20th century's best scientists and engineers, producing the transistor, laser, information theory, and Unix. Its greatness was not fully preserved with the organization itself but spread continuously through those who left. William Shockley, co-inventor of the transistor, left Bell Labs to found Shockley Semiconductor in California. In 1957, eight engineers collectively left to establish Fairchild Semiconductor. The founding teams of Intel and AMD later split from the Fairchild system. Much of what became Silicon Valley's industrial network formed along this talent flow path. Google DeepMind has not yet reached Bell Labs' conclusion. Gemini has a huge user base, and Google Cloud continues to grow. The similarity lies in top labs supported by mature commercial systems producing people who start new labs and companies outside. A lab's peak is measured by how many geniuses it gathers. The next stage is measured by what those geniuses build after leaving. This is both an achievement and a loss. For the entire tech industry, both can coexist, but for Google, this is a difficult balance to simply write off. Google's investment in Discovery Loop allows it to share future returns and provide cloud services, turning computing power expenses back into revenue. But a company can invest in external outcomes but cannot replace lost internal creativity. Dean worked at Google for 27 years. He said he only seriously considered starting a business about five weeks ago. A person who stayed nearly thirty years decided to leave in five weeks. Discovery Loop still has those familiar partners: Ghemawat responsible for systems, Vinyals for research, Quoc Le continuing automated machine learning. The division of labor hasn't changed much, nor has the research direction; they just moved out of Google's building. In his farewell letter, Hassabis said AGI is close at hand. Dean's new company also bets that AI will change scientific research. They probably still believe in the same future, just starting their separate journeys. Original Link