Developers have introduced PES Benchmark v0.2—a tool capable of effectively distinguishing live human movement from neural network-generated simulations. The method is based on biometric entropy analysis using 128-dimensional motion vectors, achieving an exceptionally high class separation rate (Cohen's d=10.4).
What Happened
PES Benchmark v0.2, a tool for detecting AI-generated motion, has been presented. The system uses a 128-dimensional motion vector extracted from 3D pose sequences and relies on mathematical constraints, such as the Nyquist limit and entropy gap, to identify signs of simulation.
Context
The MyShape Protocol project proposes a shift from static biometrics to the concept of "Proof of Continuity." Unlike traditional verification methods, this approach focuses on the dynamic continuity of presence, leveraging fundamental mathematical differences between biological and synthetic motion.
Why It Matters for the Industry
For the AI and cybersecurity industries, the emergence of specialized benchmarks sets a new standard for evaluating the "humanity" of video content. This is critical for developing defense systems against advanced deepfakes in video calls, creating robust KYC (Know Your Customer) protocols, and auditing new generative video models.
Why It Matters for Users
This technology will provide users with more reliable ways to confirm that a real person, rather than an AI model, is on the other end of a video call. This will become a key trust factor in an era of mass realistic content generation.
What Is Not Yet Known / Limitations
Full implementation in production environments (e.g., video conferencing systems) requires additional data regarding latency and API integration capabilities.
Sources
Author
Look at AI, Editorial Team