April was a defining month for Sony AI. Project Ace, the table tennis research project five years in the making, reached the cover of Nature and traveled across more than 1,800 articles in 30+ countries. Beyond Ace, our Chief Scientist, Peter Stone contributed a new framework for understanding AI companionship, returned to ICLR 2026 with a full slate of papers, and opened a window into the future of AI-powered chip design. Here is a closer look at what shaped April across the lab and beyond.
Studying the Risks and Benefits of AI Companions:
Researchers Discuss a New Framework for Understanding AI Companionship
Conversational AI is taking on a new role in people's lives. For many users, these systems are no longer just tools. They behave like friends, confidants, or even romantic partners. Recent survey data shows nearly one in five high school students say they or someone they know has had a romantic relationship with an AI. A new paper from W. Bradley Knox, Sony AI Chief Scientist Peter Stone, and colleagues offers a structured framework for understanding the potential harms and benefits of AI companionship. Read the full piece here.
Inside Project Ace:
Discover the Robot Athlete That Competes With Professional Table Tennis Players
Published on the cover of Nature, this research introduces the first real-world AI system capable of competing with and beating elite university and professional table tennis players under official rules. Project Ace combines high-speed event-based vision, low-latency control, and reinforcement learning to track and return high-spin, high-velocity shots in real time. End-to-end latency is 20.2 milliseconds, compared to roughly 230 milliseconds for elite human players. Five years of work made this possible, with progress moving from simple ball-juggling to expert-level competitive play. Ace also takes its place in a series of AI landmarks, marking the first time an AI system has achieved human expert-level play in competitive physical sport.
Read the full piece here.
To watch the short Ace film and read the research in Nature, visit: ace.ai.sony
Sony AI at ICLR 2026:
Research Roundup
Sony AI returned to the International Conference on Learning Representations (ICLR) with a slate of research that reflects where machine learning is today and where it is heading next. The contributions span multimodal embeddings, diffusion training strategies, interpretability, object-centric learning, audio tooling, video generation, and the theoretical foundations of neural reasoning. Featured work includes Concept-TRAK, which traces how visual concepts from training data influence diffusion model outputs; VIRTUE, a visual-interactive text-image embedder that lets users point to regions of interest in an image; CMT, a mid-training strategy that improves the efficiency of flow map models; and LLM2Fx-Tools, a framework that uses large language models to infer executable audio effect chains for music post-production. Many of the projects include open code, benchmarks, and demos.
Read the full roundup here.
April was a milestone month for Ace. Following its publication in Nature, the research project drew global attention, resulting in more than 1,800 articles across 60+ countries. The stories showcased the research findings and the importance of this milestone for robotics.
The Associated Press: Matt O’Brien interviewed Peter Dürr and Michael Spranger about the Ace project and indicated the importance of the research beyond table tennis. Spranger noted, “With this technology, we show that it’s actually possible to train robots to be very adaptive and competitive and fast in uncertain environments that constantly change.” Read the article here.
Reuters: Will Dunham interviewed Peter Dürr, providing readers with an in-depth look at Ace. When discussing the challenge Sony AI set out to solve with Ace, Dürr commented, “Unlike computer games, where prior AI systems surpass human experts, physical and real-time sports such as table tennis remain a major open challenge due to their requirements for fast, precise and adversarial interactions near obstacles and at the edge of human reaction time… The project's goal was not only to compete at table tennis but to develop insights into how robots can perceive, plan and act with human-like speed and precision in dynamic environments." Read the article here.
NZZ: Gioia da Silva spoke with Peter Dürr to learn more about the Ace project, how the robot was trained using reinforcement learning, and what it means for the future of robotics. Durr notes that the robot “...learned to play table tennis without any human role model, solely through trial and error." Read the article here.
Some other notable stories about the research…
Sony AI in Silicon Semiconductor:
Lorenzo Servadei on the Future of AI-Powered Chip Design
Lorenzo Servadei, Head of AI for Chip Design at Sony AI, was featured in the latest issue of Silicon Semiconductor Magazine. The interview explores how AI is reshaping semiconductor design from initial concept through fabrication. Lorenzo discusses two recent research contributions from his team.
GENIE-ASI is a training-free, LLM-based method for analog subcircuit identification. It produces human-readable reasoning alongside executable Python from just a few examples.
Schemato translates netlists into human-interpretable schematics that engineers can load and simulate in LTSpice. The conversation also covers the trade-off between automation and human oversight, the role of multi-physics co-optimization, and where AI can push chip design over the next five years.
Coming Soon: Sony AI at ICASSP 2026
The IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) is one of the longest-running venues for advances in audio, speech, and signal processing. This year's conference takes place May 4–8, 2026 at ICASSP 2026 in Barcelona, Spain. Sony AI will present 11 accepted papers spanning music understanding, generative audio, audio-visual alignment, and data quality. We will return with a full roundup, paper deep-dives, and author commentary as the conference approaches.
In the meantime, dive into our work from last year's ICASSP: Unlocking the Future of Video-to-Audio Synthesis: Inside the MMAudio Model.
Coming Soon: Sony AI at CVPR 2026
The Computer Vision and Pattern Recognition conference (CVPR) is one of the most significant annual gatherings in computer vision research. It brings together work that shapes how machines perceive, reconstruct, and generate visual information. This year's conference takes place June 3–7, 2026 at the Colorado Convention Center in Denver. CVPR 2026 will feature six accepted papers from Sony AI and its collaborators, spanning generative modeling, physical AI, 3D scene understanding, and robust perception. We will return with a full roundup and paper deep-dives as the conference approaches.
In the meantime, dive into our work from last year's CVPR: Research That Scales, Adapts, and Creates: Spotlighting Sony AI at CVPR 2025.
April demonstrated the breadth of research underway at Sony AI; from physical intelligence on the table tennis court to the foundations of learning representations, and the future of chip design. With ICASSP 2026 and CVPR 2026 ahead, May and June will bring new research deep-dives and conference roundups.
Connect with us on LinkedIn, Instagram, or X, and let us know what you’d like to see in future editions. Until next month, keep imagining the possibilities with Sony AI.