Advancing AI: Highlights from August

August was a month of sound, circuits, and shared creativity. From new approaches in audio search to breakthroughs in analog design, Sony AI continued advancing research that makes AI more adaptive, interpretable, and supportive of human ingenuity. Here’s a look at what happened this month across our teams, labs, and global stages.

Behind the Sound: How AI Is Enhancing Audio Search

Finding the right sound for sound designers can often mean hours of manual searching or relying on imperfect labels. That friction sparked a research collaboration with Audiokinetic, resulting in Similar Sound Search (SSS), a new audio-to-audio and text-to-audio feature. Now available in the Wwise Beta as of August 19, 2025—Wwise is an industry-standard audio middleware solution developed by Audiokinetic.

In August, Sony AI shared how we were part of this process of supporting creators with audio-to-audio and text-to-audio search capabilities. Instead of limiting sound discovery to keywords or metadata, this collaboration makes it possible to:

-Search for a sound by providing another sound as the query.
-Record, mimic, or reference audio snippets to locate similar assets.
-Free creators from relying solely on descriptive tags or subjective labels.

By pairing advanced AI models with professional sound libraries, we’re helping sound designers find what they need faster and with greater creative freedom.

Read the full blog: Behind the Sound: How AI Is Enhancing Audio Search
Read the Press Release: Sony AI and Audiokinetic Partner to Create First AI-Powered Text-to-Audio and Audio-to-Audio Sound Effect Search Tool for Professional Production

Why the Hardest AI Problems Still Matter

Our latest blog takes you inside Sony AI’s Reinforcement Learning Team, where researchers like James MacGlashan, Harm van Seijen, Varun Kompella, and Chief Scientist Peter Stone share why RL remains essential to advancing AI.

From powering GT Sophy to shaping the future of robotics, RL is about more than predictions—it’s about learning to act, adapt, and align with human values.

Read the full post: Sony AI’s Deep RL Team on Why the Hardest Problems Still Matter

MLCAD 2025: LLMs in Analog Design

Analog circuits may be mighty, but their design process has historically resisted automation. At MLCAD 2025, Sony AI will present two complementary approaches that embed large language models into the analog workflow:

GENIE-ASI: Learns how to identify analog subcircuits from just a handful of examples, then generates reusable Python code to automate detection at scale.

Schemato: A fine-tuned LLM that converts netlists into human-readable schematics with high connectivity and fidelity.

Together, these tools address two long-standing challenges—identifying a circuit’s functional building blocks and making designs easily interpretable—pointing toward a future where AI supports engineers with scalable, trustworthy tools.

Stay tuned for more details!

This month, our collaboration with Audiokinetic sparked conversations across gaming, sound, and beyond. The media explored how Similar Sound Search could shape creative possibilities in each of these areas.

• GamesBeat | Sony AI and AudioKinetic enable audio-to-audio and text-to-audio search for devs
• Sound on Sound | Audiokinetic & Sony AI reveal Similar Sound Search

Some of our researchers and engineers also recently sat down with Nikkei to discuss advances in Sony AI’s gaming and computer vision work. From GT Sophy to our Vision Foundation Model, the conversations highlighted how our projects have made an impact.

• Nikkei Robotics | SonyAI develops Lightweight CV foundation model, achieving Top-Level Performance in Over 10 CV tasks
• Nikkei XTech | Sony AI develops well-behaved racing AI to improve both people and driving

On the Horizon: Where to Find Us Next

August is a wrap, but the momentum continues into the fall:

MLCAD 2025 (Sept 8–10 | Santa Cruz, Calif.)
Sony AI is a Silver sponsor with two accepted papers. Our researchers will continue conversations around GENIE-ASI and Schemato—two projects demonstrating how AI can augment, rather than replace, expert reasoning in analog design. Stay tuned for our upcoming blog post which will deep-dive into the research!

ISMIR 2025 (Sept 21-25 | Daejeon, Korea)
This year, we will be back with new, soon-to-be-announced audio and music-based research but if you’d like to take a look at our past years work at ISMIR, check out the following research below:

• -SpecMaskGIT: Masked Generative Modeling of Audio Spectrograms for Efficient Audio Synthesis and Beyond – Sony AI
• -Towards Assessing Data Replication in Music Generation with Music Similarity Metrics on Raw Audio – Sony AI
  
• -Automatic Piano Transcription with Hierarchical Frequency-Time Transformer – Sony AI

CSCW 2025 Workshop: Responsibly Training Foundation Models
October 18, 2025 | Hybrid

Sony AI is co-organizing a full-day workshop at CSCW 2025 on “Responsibly Training Foundation Models: Actualizing Ethical Principles for Curating Large-Scale Training Datasets in the Era of Massive AI Models.” Sony AI researchers Morgan Klaus Scheuerman, Jerone T. A. Andrews, and Alice Xiang will be co-organizing the workshop, alongside collaborators from Stanford, Trinity College Dublin, University of Michigan, King’s College London, Arizona State University, and others.

This hybrid event will bring together researchers and practitioners from across the globe, and across disciplines, to address the unique challenges of curating datasets for foundation models. Building on CSCW’s tradition of interdisciplinary exchange, the workshop will explore themes around dataset composition, curation processes, and release practices — with an emphasis on cultural, social, legal, and ethical considerations.

Outcomes will include a conceptual framework to guide more responsible large-scale dataset curation, with discussions ranging from ethical labor practices in annotation to the “right to be forgotten” in AI training data.

Learn more: https://responsible-data-workshop.github.io/cscw2025/

Sony AI at ICCV 2025
October 19–23, 2025 | Hawai‘i Convention Center, Honolulu, Hawai‘i

Sony is proud to be a Silver Sponsor of the International Conference on Computer Vision (ICCV 2025). Sony AI will also be on site with a booth and is excited to share that our researchers have 7 accepted papers at this year’s conference.

Stay tuned for more details on the research highlights in the coming months.

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.