Guests:
- Yanjing Li, University of Chicago, USA
- Luciano Ost, Loughborough University, UK
Host:
- Maksim Jenihhin, TalTech, EE
From cosmic rays and scary silent data corruptions to self-healing chips, we discuss ensuring reliable hardware in the AI era.
- What are the Silent Data Corruptions in simple terms? What causes them?
- Why and when should we care about cosmic rays coming from the Sun?
- How does the FIDELITY framework by UChicago help analyse fault resiliency of DNN accelerators and discover new vulnerabilities?
- Why cross-layer reliability assessment for edge AI, and when shall we “torture” chips with radiation beams?
- Is research in the USA, close to the top corporations, different from the research we do in Europe?
- What advances might the next 5 years bring to make our computing systems even more reliable?
Recorded: April 2025, TalTech, Tallinn
Published: May 2025
