Seth Dobrin, IBM | IBM Think 2021
summary
IBM describes trust as five concrete pillars: transparent, fair, explainable, robust, privacy. Claims include explainable black‑box models today and open source AI Fairness 360 toolkits for bias detection. Architecture examples: models train on cloud GPUs, deploy to mainframes for inference to meet half‑millisecond SLAs for credit‑card fraud, and use federated learning to synchronize ~50 distributed model instances worldwide. Use cases given: mortgage and recidivism risk scoring, real‑time fraud detection, smart factories and edge inference.
AI-written summary · about
