ProtoPilot: A Full-Chain Agent System That Learns from Failure <\/p>\n
ProtoPilot\u00a0is a self-evolving multi-agent system that covers the entire experimental lifecycle:
Design2Protocol \u2192 Protocol2Code \u2192 Device Execution \u2192 Wet-Lab Feedback <\/p>\n
It learns from failure. When a PCA assembly step failed, ProtoPilot diagnosed the issue (antibiotic resistance screening failure) and autonomously regenerated a corrected protocol\u2014proving true Physical AI is here. <\/p>\n
On\u00a0ProtocolQA, one of the most representative public benchmarks for evaluating AI experimental reasoning capabilities (built by AI4S leader Future House): <\/p>\n GPT-5.6-sol\u00a0scored\u00a043.5% <\/p>\n Human expert\u00a0level stands at\u00a054% <\/p>\n ProtoPilot\u00a0achieved\u00a052.38%\u00a0\u2014approaching expert-level performance! <\/p>\n BioLab Bench: The First Real-Task Evaluation System for Life Science Agents <\/p>\n BioLab Bench\u00a0sets a new industry standard. It’s the first evaluation system that measures not just “correct answers,” but whether an agent can actually execute tasks on real automation equipment. <\/p>\n Key features include: <\/p>\n Real-World Task Coverage: BioLab Bench spans from fundamental operations to complex multi-step workflows, stratified across three difficulty levels (L1\u2013L3). <\/p>\n Full-Chain Assessment: Rather than merely checking whether an agent generates a plausible protocol, BioLab Bench evaluates each step\u2014intent interpretation, protocol design, device-agnostic SOP generation, device-specific SOP translation, machine code production, and successful execution gate verification. <\/p>\n Cross-Device Transferability: The benchmark can be deployed on different automated laboratory platforms to test whether an AI agent can comprehend experimental tasks and generate executable actions adapted to varying hardware configurations, thus assessing cross-device generalization capability. <\/p>\n Story Continues <\/p>\n Toward 7\u00d724 Unattended Smart Laboratories <\/p>\n Moving forward, BioAgents will no longer improve solely through text-based training. Instead, through the PhysicalAI experimental loop, they will continuously accumulate real research tasks, automation operations, expert validations, failure cases, and wet-lab feedback. This massive corpus of physical experimental data will enable BioAgents to develop integrated reasoning, execution, and validation capabilities\u2014ultimately powering 7\u00d724 unattended intelligent laboratories. <\/p>\n Built on a Decade of Synergy Between AI and Biology <\/p>\n MGI’s exploration of AI dates back to 2019. In 2025, the team led by Dr. Yang Meng, Chief AI Officer of MGI, in collaboration with Professor Nattiya Hirankarn from Chulalongkorn University, published a paper in\u00a0Nature Biomedical Engineering\u00a0introducing “PrimeGen”\u2014a dry\u2013wet collaborative multi-agent system that integrated primer design, experimental validation, and automated workstation execution into a closed-loop workflow. <\/p>\n In April 2026, MGI established Genoria AI as a dedicated subsidiary focused on AI for Science (AI4S), with a mission to build dry\u2013wet closed-loop infrastructure for the life sciences. <\/p>\n The new Physical AI initiative builds on MGI’s unique strengths in hardware-native advantages\u00a0with deep integration across its automation platforms, and real-world deployment expertise\u00a0gained from over 3,800 users globally. <\/p>\n “It reflects a different path from the pure compute race. While leading AI companies rely on scale compute to push the capabilities of general-purpose models, we take a different approach. Through agent scaling and closed-loop data engineering, we organize real-world tasks, device constraints, expert feedback, and wet-lab results into a training ground where AI continuously evolves.” Noted Dr. Yang Meng, now serving as CEO in Genoria AI. <\/p>\n About Genoria AI <\/p>\n Genoria AI, a subsidiary of MGI Tech, is an AI4S company building self-evolving labs to scale agentic discovery. Learn more: https:\/\/www.genoria.ai\/en<\/a> <\/p>\n About Shanghai Artificial Intelligence Laboratory <\/p>\n The Shanghai AI Laboratory was officially unveiled at the World AI Conference (WAIC) in July 2020 and positioned as a national-level new-type research institute. Our vision is to build a world-class AI laboratory, with pioneering contributions on original theories and key technologies. By gathering top talents from around the world and creating a stimulating and collaborative research environment, the Laboratory aims to conduct original and influential research, make fundamental contributions to basic theories, significantly advance the state of the art, and create remarkable impact on industry, healthcare, and education. Learn more: https:\/\/www.shlab.org.cn\/<\/a> <\/p>\n<\/a> ProtoPilot\u00a0achieved\u00a052.38% on ProtocolQA <\/p>\n
<\/a> BioLab Bench <\/p>\n