Google AI is a subsidiary of Google DeepMind dedicated to advancing Artificial intelligence research and developing related products. The division was announced at Google I/O 2017 by then-CEO Sundar Pichai, consolidating the company's AI efforts under a single umbrella. Over time, it expanded its research facilities to locations including Zurich, Paris, Israel, and Beijing, reflecting a global commitment to AI development.
In 2023, Google AI became part of a major reorganization that merged its internal machine learning unit, Google Brain, with DeepMind, a UK-based company acquired in 2014 that had previously operated separately. This merger created a unified entity known as Google DeepMind, and elevated Jeff Dean, the head of Google AI, to the position of chief scientist at Google. The reorganization aimed to accelerate progress in AI by combining research talent and resources.
History and Formation
The origins of Google AI trace back to earlier research groups within Google, including Google Brain, which was established in 2011. Google Brain focused on deep learning and neural networks, contributing foundational work in areas such as large language models and transformers. The formal announcement of Google AI at Google I/O 2017 signaled a strategic push to coordinate these efforts and bring AI products to a broader audience.
Under the leadership of Jeff Dean, Google AI pursued a dual mission: advancing fundamental research and applying AI to real-world products. The division collaborated with academic institutions and published extensively, while also developing internal tools that would later become widely used in the industry.
Merger with DeepMind
The 2023 merger between Google Brain and DeepMind was a pivotal moment for Google AI. DeepMind, known for breakthroughs such as AlphaGo and protein structure prediction, brought a strong research culture and a focus on long-term AI safety. The combined entity, Google DeepMind, operates under the leadership of Demis Hassabis, co-founder of DeepMind, with Jeff Dean serving as chief scientist.
This restructuring allowed for closer collaboration between teams working on generative AI, reinforcement learning, and other advanced topics. It also clarified the organizational structure for external partners and researchers, making Google DeepMind the primary hub for AI innovation within Alphabet.
Projects and Products
Google AI has been associated with a wide range of projects, from research tools to consumer-facing applications. One notable initiative is Google Vids, an AI-powered video creation tool designed for workplace use. Another is Google Assistant, a virtual assistant that has been developed by Google AI since 2023, integrating AI capabilities into everyday tasks.
A significant technical contribution is the provision of cloud-based TPUs (tensor processing units) for machine learning development. The TPU research cloud offers free access to a cluster of cloud TPUs for researchers engaged in open-source machine learning research, lowering the barrier to entry for those working on large-scale models.
TensorFlow
TensorFlow, a machine learning software library created by Google Brain in 2015, is one of the most widely used tools in the AI community. It provides a flexible framework for building and deploying models, supporting everything from research prototypes to production systems. TensorFlow has been instrumental in spreading deep learning techniques across industries.
Magenta
Magenta is a deep learning research team within Google AI that explores the role of machine learning as a tool in the creative process. The team has released numerous open-source projects that allow artists and musicians to extend their workflows using AI. With Magenta, musicians and composers can generate high-quality music at a lower cost, making it easier for new artists to enter the industry.
Sycamore
Sycamore is a 54-qubit programmable quantum processor developed by Google. While not strictly an AI project, it has implications for future computing capabilities, including potential applications in machine learning. The processor achieved a milestone in 2019 when it demonstrated quantum supremacy, performing a specific task faster than the most powerful classical supercomputers at the time.
LaMDA
LaMDA (Language Model for Dialogue Applications) is a family of conversational neural network models designed to improve natural language understanding in dialogue. LaMDA was a precursor to later models and contributed to the development of conversational AI systems.
Datasets in Under-Represented Languages
Google AI has also focused on creating datasets in under-represented languages to facilitate the training of AI models in these languages. This effort aims to make AI technologies more inclusive and accessible globally, addressing the bias that can arise from training primarily on English-language data.
Former Projects
Some projects that were initially part of Google AI have since been transferred to Google DeepMind or discontinued. Bard, a chatbot based on the Gemini model, was no longer developed by Google AI as of February 8, 2024, when it was merged into the Gemini brand and moved to Google DeepMind. Similarly, Duet AI, a Google Workspace integration that could generate text or images, was also transferred to Google DeepMind on the same date.
Crowdsource, a crowdsourcing platform developed by Google, was another former project. It aimed to improve Google services through data donated by users for training various algorithms. The platform allowed users to contribute translations, transcriptions, and other data to help improve AI models.
Ethical Considerations and Controversies
In March 2019, Google announced the creation of the Advanced Technology External Advisory Council (ATEAC), comprising eight members: Alessandro Acquisti, Bubacarr Bah, De Kai, Dyan Gibbens, Joanna Bryson, Kay Coles James, Luciano Floridi, and William Joseph Burns. The council was intended to provide external guidance on the ethical implications of Google's AI technologies. However, following objections from a large number of Google staff to the appointment of Kay Coles James, who had made controversial statements on LGBTQ issues, the council was abandoned within one month of its establishment.
In February 2025, Alphabet removed guidelines from its public AI ethics policy that formerly ruled out applying AI technology to applications that were "likely to cause harm." Google published a blog post defending the change, arguing that the previous policy was too restrictive and that AI should be developed with appropriate safeguards rather than blanket prohibitions. This move sparked debate within the AI community about the balance between innovation and ethical responsibility.
Impact and Legacy
Google AI has had a profound impact on the field of artificial intelligence, both through its research contributions and its open-source tools. TensorFlow alone has become a standard in the industry, used by startups and large enterprises alike. The division's work on transformers and large language models has influenced the development of subsequent AI systems, including those from competitors like OpenAI and Anthropic.
The merger with DeepMind in 2023 positioned Google to compete more effectively in the rapidly evolving AI landscape. By combining the strengths of Google Brain and DeepMind, Google DeepMind has continued to push the boundaries of what is possible in AI, from scientific discovery to creative applications.
Future Directions
As of 2025, Google DeepMind continues to develop the Gemini family of models, which represent the latest in generative AI technology. These models are integrated across Google's product ecosystem, including search, cloud services, and consumer devices. The focus on responsible AI development remains a priority, though the company's evolving ethics policies suggest a willingness to adapt to changing technological and competitive pressures.
Google AI's legacy is one of innovation and influence, shaping the tools and techniques that define modern AI. Its contributions to open-source software, research, and product development have left an indelible mark on the field, and its reorganization into Google DeepMind ensures that this impact will continue for years to come.