Mustafa Suleyman, a cofounder of DeepMind, previously revolutionized artificial intelligence by developing techniques that helped computers solve complex problems. His company, acquired by Google in 2014, gained fame with AlphaGo, a program that defeated the world’s best Go player in 2016. Now, as the CEO of Microsoft AI, Suleyman is introducing a new AI breakthrough aimed at enhancing user interaction with technology. Suleyman's recent advancements include updates to Microsoft’s AI assistant, Copilot, which now features a humanlike voice, the capability to see user screens, and improved reasoning abilities. He aims to rekindle users' appreciation for PCs by designing AI companions that engage in more persistent, supportive relationships with users. In an interview, Suleyman elaborates on Copilot's new vision, highlighting its ability to understand and respond to user queries in real-time. He believes AI should provide emotional support and create lasting interactions akin to companionship.

The Copilot Vision feature allows users to ask questions about what they observe on their screens without needing to type, thus easing digital navigation. Despite comparisons to previous Windows assistants like Clippy, Suleyman reassures users that privacy is prioritized, as Copilot Vision does not save information after sessions. He introduces the "Think Deeper" feature, built on OpenAI’s model, which enhances the AI's problem-solving capabilities by facilitating side-by-side comparisons and deeper reasoning. Suleyman confirms Copilot is currently available to Microsoft employees, with ample feedback being collected. Envisioning the future, he discusses the potential for AI agents that can autonomously perform tasks such as making purchases. Though developing fully capable agents is still a work in progress, he is optimistic about a gradual advancement in the technology. Key challenges revolve around ensuring trust in AI interactions, particularly regarding privacy and security. Suleyman emphasizes the importance of establishing clear boundaries for AI agents to foster reliable and trusted user experiences as they evolve to assume more complex tasks.