AI, Algos, and Bias

Question 1

define AI

Answer 1

AI models are typically built using machine learning which is the process of finding patterns in data

1. get lots of data
2. feed teh data into a machine learning algo which uses optimization to find algos
3. the algo determines patterns and produces a model for decision making through training
4. the choice of which criterion to optimize is made by the programmer
5. when model has a suitable high accuracy rate in predictions, it’s ready to use

Question 2

describe how AI affects our lives

Answer 2

1. online dating, healthcare, and price we pay, qualifying for loan, housing orwelfare benefits, ads delievered online
2. early warning of mental health problems, potential tax fraud, length of criminal sentences, eligibility for parole
3. many other things

Question 3

know some examples of bias in AI

Answer 3

1. Amazon: AI hiring process which can be cheated by applicants who put white text in margins that are machine readable but invisible to human such as Harvard
2. facial recognition system

Question 4

solutions to AI and robotic bias

Answer 4

1. transparent data
2. encourage formation of multidisciplinary teams

Question 5

how to govern algos

Answer 5

1. transparency
2. auditability
3. commitment to due process: able to appeal as an individual even an algo is considered fair