MIT News breed · Wetenschap

MIT-IBM Lab Fosters Transition from AI/Quantum Research to Industry Application

The MIT-IBM Computing Research Lab has facilitated the transition of former MIT graduate students and postdocs into impactful roles at IBM, enabling them to bridge theoretical research with real-world business applications in AI and quantum computing.

Srinivasan Arunachalam, Zhang-Wei Hong, and Irene Ko, all formerly associated with MIT and now at IBM, leveraged their experience at the MIT-IBM Computing Research Lab to translate novel research into practical business solutions. Their work spans quantum machine learning, reinforcement learning, AI agents, and trustworthy AI.

Hong, focusing on reinforcement learning, developed techniques to improve AI grounding for realistic applications and enhance reward feedback, which he applied to robotics, large language models (LLMs), and reinforcement learning for science. He is now developing infrastructure for IBM's agentic framework, aiming to create a system where models can self-evolve at deployment time.

Ko's research in trustworthy AI, funded by MIT-IBM, aligns with industry standards for robust, accurate, and fair AI. She developed a lightweight vLLM inference engine plugin framework, vLLM Hook, to analyze internal model signals for safety, offering a bridge between development and deployment in trustworthy AI.

Arunachalam, exploring quantum machine learning, focused on problems implementable on near-term quantum devices. His work, shaped by the MIT-IBM connection, led to papers on Hamiltonian learning and quantum kernels, providing theoretical guarantees and evidence for quantum advantages.

All three researchers emphasize the value of interdisciplinary connections and collaborations, like that fostered by the MIT-IBM lab, in developing practical applications that demonstrate the real-world impact of their research.

AI-samenvatting op basis van de bron.