• Posts
  • Microsoft hires DeepMind co-founder Mustafa Suleyman to run new consumer AI unit

Microsoft hires DeepMind co-founder Mustafa Suleyman to run new consumer AI unit

Also: DeepMind and Liverpool FC develop AI to advise on football tactics


In today's newsletter, we delve into Microsoft's strategic hiring of Mustafa Suleyman to head its new consumer AI unit, aiming to enhance its AI products' integration and market position. In sports technology, DeepMind collaborates with Liverpool FC to develop TacticAI, an AI model aimed at improving football tactics, showcasing the growing influence of AI in sports. Apple steps up in the AI race with its new MM1 model, signaling a stronger push into generative AI, potentially altering future product offerings and privacy approaches. A Harvard study highlights the nuanced impacts of AI on radiology, challenging the notion that AI tools uniformly enhance diagnostic accuracy. In health and fitness, Google's new AI project aims to provide personalized advice through Fitbit, marking an advancement in AI-driven health recommendations. In gaming, Ubisoft's exploration with AI-powered NPCs promises a future of more interactive and immersive gaming experiences. Lastly, NeuReality's recent funding bolsters its efforts to revolutionize AI data center efficiency, highlighting the ongoing rapid development and application of AI across various sectors.


  • 🤝 Microsoft hires DeepMind co-founder Mustafa Suleyman to run new consumer AI unit

  • ⚽️ DeepMind and Liverpool FC develop AI to advise on football tactics

  • 👀 Apple’s MM1 AI Model Shows a Sleeping Giant Is Waking Up

  • 🏥 AI Tools Can Hurt Radiologist Performance, Harvard Study Finds

Microsoft has appointed Mustafa Suleyman, co-founder of DeepMind and CEO of the AI startup Inflection, to lead a new consumer AI unit, Microsoft AI. This strategic move aims to consolidate Microsoft's significant advancements and investments in the generative AI sector, including a $13 billion investment in OpenAI, creators of ChatGPT, and the integration of AI technologies into Microsoft products. Suleyman's role will involve merging consumer-facing products like Microsoft’s Copilot, Bing, Edge, and GenAI into one cohesive team, enhancing Microsoft's position in the competitive AI market. The hiring signifies Microsoft's intent to innovate further in AI technologies, impacting various products, including the Windows operating system and the Bing search engine. Additionally, Microsoft will absorb a substantial portion of Inflection’s staff, bringing in top AI engineers, researchers, and model builders. This shift highlights Microsoft's commitment to leading AI technology development and deployment, ensuring that the benefits of AI reach a broad audience while addressing safety and responsibility concerns.

DeepMind has partnered with Liverpool FC to create TacticAI, an innovative AI model designed to optimize football strategies, particularly focusing on corner kicks. This collaboration aims to leverage AI to improve the tactical aspects of the game by analyzing data from thousands of corner kicks to predict outcomes and suggest effective plays. TacticAI, which was developed over a three-year research collaboration, employs historical data from Premier League matches to forecast which player is likely to first contact the ball post-corner. The model's predictions are impressively accurate, placing the actual receiver in its top three choices 78% of the time. Coaches can utilize this AI tool to develop tailored attacking or defensive strategies, enhancing or diminishing the likelihood of a certain player receiving the ball or a team making a successful goal attempt. Despite its capabilities, the AI is not meant to replace human coaches but to augment their strategic planning, allowing them more creative freedom. The research, while focused on football, is seen as having broader implications, potentially improving the understanding of human psychology and decision-making under uncertainty, aspects relevant to advancing AI capabilities in various fields.

Apple has recently unveiled its AI model named MM1, marking a significant step in the tech giant's journey into generative artificial intelligence. The MM1 model is versatile, handling both text and image analysis, showcasing capabilities similar to those of OpenAI's ChatGPT and Google's Gemini. This development indicates Apple's move towards integrating advanced AI in its offerings, despite its previous reticence in the AI-generated content space. MM1, a multimodal large language model, represents Apple's foray into the next generation of foundational AI models, suggesting potential new directions for products and services. Although smaller in scale compared to other tech giants' models, MM1's design allows for easier experimentation and refinement by Apple engineers. Detailed in a research paper, the model's training techniques and multimodal nature hint at future applications such as an advanced AI assistant or enhanced capabilities for Siri, aligning with Apple's emphasis on user privacy and on-device processing. As Apple continues to invest in AI, MM1 could play a crucial role in maintaining the company's competitive edge in technology and privacy, leveraging its control over software and hardware integration.

A Harvard Medical School study examining the impact of AI tools on radiologist performance has yielded surprising results, revealing that AI assistance does not universally enhance diagnostic accuracy. Involving 140 radiologists in tests using 15 chest X-ray diagnostic tasks, the study found no consistent improvement in performance with AI assistance across different levels of expertise or years of experience. Intriguingly, the study highlights that AI tools influenced the treatment decisions for 324 patients involved, underscoring the varied impacts of AI on medical professionals' decisions. The research underscores the complex dynamics between radiologists and AI tools, showing that AI can both help and hinder different individuals. This variability emphasizes the importance of understanding the diverse interactions between clinicians and AI systems. Despite the mixed outcomes, the study suggests not abandoning AI in medical settings but calls for a deeper exploration into optimizing human-AI collaboration.

🛠️ AI tools updates

Google is advancing into the AI healthcare sector with the development of a new AI designed to provide health and fitness advice through Fitbit devices. This initiative, part of Google's broader effort to create AI health algorithms and robotic doctor tools, aims to harness the latest large language model technologies for personalized user guidance. The AI, developed in collaboration between Fitbit and Google Research, will analyze data from Fitbit and Pixel devices to offer customized recommendations, particularly focusing on sleep patterns and workout intensities. Google's Global Lead of AI, Yossi Matias, highlighted the potential of this model to bring actionable insights directly to users, enhancing their health management. While the project's release date remains undisclosed, the AI has shown promise in sleep medicine certification tests and will operate on Google's Gemini AI model. This health-focused AI, training on anonymized real human data, is part of Google's larger suite of health-oriented innovations, including AI tools for chest X-ray analysis and partnerships aimed at improving cancer screening and skin condition diagnosis.

Ubisoft, in collaboration with InWorld and Nvidia, is innovating in the gaming industry with its experimental "Neo NPCs," a new kind of AI-powered non-player characters designed to create more immersive and interactive gaming experiences. These advanced NPCs, showcased in a prototype phase, engage players in nuanced conversations, adapt to individual player responses, and contribute to evolving game narratives, thereby enhancing emotional and strategic depth within games. Despite being in the early stages of development and exhibiting some limitations such as response delays and occasional awkwardness, these NPCs represent a significant step toward more dynamic and responsive game environments. Ubisoft's initiative reflects a broader trend towards integrating AI in gaming to make virtual worlds more intelligent and adaptive to player actions, promising a future where games are more personalized and engaging.

💵 Venture Capital updates

NeuReality, an Israeli startup, has successfully raised $20 million from various investors including the European Innovation Council (EIC) Fund, Varana Capital, Cleveland Avenue XT, Hi-Tech, and OurCrowd. This funding boosts NeuReality's total investment to $70 million, enabling the company to expand the deployment of its NR1 AI Inference Solution. This move is set to transition NeuReality from its initial deployment phase to broader market growth, particularly in the domain of generative AI applications. The company's innovative NR1 AI Inference Solution, built around the 7nm AI inference server-on-a-chip called NR1 NAPU, developed with TSMC, offers a more efficient system architecture. This architecture allows companies to operate generative AI applications and large language models (LLMs) more economically, without the need for heavy investment in scarce and often underutilized GPUs.

🫡 Meme of the day

⭐️ Generative AI image of the day