Which of the following is a viable method to reduce bias when humans are deployed to label data for training an AI?Group of answer choicesUse only a single annotator for one set of data to reduce inconsistency.Use a large set of annotators with diverse demographic characteristics.Use another AI to suggest labels and humans adjust them if needed.Use only publicly available data, avoiding private and sensitive information.

Which of these are good practices for addressing bias in AI? (Select all that apply)1 pointTechnical solution such as “zeroing out” biasSystematic auditing processes to check for biasUsing more inclusive/less biased dataUsing an adversarial attack on the AI system to change its outputs to be less biased

Question 5In AI, what does bias do?  1 pointAugments human intelligence Solves problems faster Identifies and addresses socio-technical issues raised by AI  Gives systematic disadvantages to certain groups or individuals

AI can exhibit biased behaviour becauseAI cannot exhibit biased behaviourIt intentionally prefers certain outcomesIt is programmed to be biased by developersIt can replicate biases present in the training data

Question 6What is a first step toward mitigating bias in AI?  1 pointAssembling diverse teams  Putting into place a governance structure that works at scale  Developing different rules for different risks  Designating a lead AI ethics official

As a data scientist, you are tasked with identifying bias in training data in an effective way. Which of the tools below can be used to identify bias in data?TensorFlow Data Validation, TensorFlow Serving.TensorFlow Data Validation, What-if Tool.TensorFlow Datasets, What-if Tool.TensorFlow Data Validation, TensorBoard.