Explaining Decisions from Vision Models and Correcting them via Human Feedback

Channel:
Microsoft Research
Subscribers:
344,000
Published on ● Video Link: https://www.youtube.com/watch?v=Z-dWhly8ybE


Duration: 1:07:04
803 views
22

Deep networks have enabled unprecedented breakthroughs in a variety of computer vision tasks. While these models enable superior performance, their increasing complexity and lack of decomposability into individually intuitive components makes them hard to interpret. Consequently, when today's intelligent systems fail, they fail spectacularly disgracefully, giving no warning or explanation.

Towards the goal of making deep networks interpretable, trustworthy and unbiased, in my talk I will present my work on building algorithms that provide explanations for decisions emanating from deep networks in order to —
• understand/interpret why the model did what it did,
• correct unwanted biases learned by AI models, and
• encourage human-like reasoning in AI.

See more at https://www.microsoft.com/en-us/research/video/explaining-decisions-from-vision-models-and-correcting-them-via-human-feedback/



Other Videos By Microsoft Research


2020-06-14Local Context Normalization: Revisiting Local Normalization
2020-06-10Supervised Deep Hashing for Efficient Audio Retrieval
2020-06-10Transparency and Intelligibility Throughout the Machine Learning Life Cycle
2020-06-10Machine Learning and Fairness Webinar
2020-06-10Consumer Brain-Computer Interfaces: From Science Fiction to Reality
2020-06-10Highly Conductive Flexible Sensor Integrated With Personal Devices For Practical Bio-Signal Measure
2020-06-08Microsoft Build 2020: Kevin Scott keynote with Lila Tretikov
2020-06-03Harvesting randomness, HAIbrid algorithms and safe AI with Dr. Siddhartha Sen | Podcast
2020-06-03Provably efficient reinforcement learning with Dr. Akshay Krishnamurthy | Podcast
2020-06-01What ‘bhasha’ do you want to talk in? With Kalika Bali and Dr. Monojit Choudhury | Podcast
2020-05-26Explaining Decisions from Vision Models and Correcting them via Human Feedback
2020-05-26Auditing Outsourced Services
2020-05-26MSR Distinguished Lecture Series: First-person Perception and Interaction
2020-05-26Large-scale live video analytics over 5G multi-hop camera networks
2020-05-26Kristin Lauter's TED Talk on Private AI at Congreso Futuro during Panel 11 / SOLVE
2020-05-19How an AI agent can balance a pole using a simulation
2020-05-19How to build Intelligent control systems using new tools from Microsoft and simulations by Mathworks
2020-05-13Diving into Deep InfoMax with Dr. Devon Hjelm | Podcast
2020-05-08An Introduction to Graph Neural Networks: Models and Applications
2020-05-07MSR Cambridge Lecture Series: Photonic-chip-based soliton microcombs
2020-05-07Multi-level Optimization Approaches to Computer Vision


Tags:
computer vision
Deep networks
AI models
human-like reasoning in AI
vision models
Ramprasaath Selvaraju
Ece Kamar
Microsoft Research