class: center, middle, inverse, title-slide # Review, and AI Ethics Discussion 1 ### Ken Arnold ### 2021-02-17 --- ## Logistics * Attendance * Piazza: Live Q&A during class today * Forgot to post Week 1 Q&A; see website * Lab 1 walkthrough videos posted * 1-on-1s * Pace? --- ## [Obituary](https://www.nytimes.com/2021/02/09/science/arianna-wright-dead.html): Dr. Arianna Rosenbluth (1927-09-15 - 2020-12-28) .pull-left[ <img src="https://static01.nyt.com/images/2021/02/16/obituaries/05rosenbluth1/merlin_183269556_7d98de1d-baac-422a-8cf0-163383ac27f7-superJumbo.jpg" height="100%" style="display: block; margin: auto;" /> ] .pull-right[ * PhD in Physics from Harvard at age 21 * Programmed the first implementation of the Metropolis algorithm, in machine code (1953) ] --- ## The Metropolis algorithm Goal (for most ML applications): run a generative model backwards. .pull-left[ * You have observations * You propose a theory: a program that generates those observations * But the program depends on unknown constants * You need: good values of those constants ] .pull-right[ * documents (collections of words) * generate documents by sampling words from topics * which words are in which topics? * algorithm: Latent Dirichlet Allocation ] (Imagine running your image classifier backwards: tell it "dog" and it should generate an image of a dog.) --- > Instead of choosing configurations randomly, then weighting them with exp(−E/kT), we choose configurations with a probability exp(−E/kT) and weight them evenly. <!--  --> <https://mbjoseph.github.io/posts/2018-12-25-animating-the-metropolis-algorithm/> --- class: center, middle ## Open Q&A --- ## AI Ethics Examples (We'll do this in the class OneNote) --- ## Aspects of AI that affect ethics * AI needs data (privacy) * ...and empowers organizations that aggregate data * AI needs lots of computation * AI's outputs affect people * predictions and decisions (bias, discrimination, gig labor, ...) * how we perceive each other (and ourselves) * how we perceive ideas (misinformation) * AI does things that people once did * AI systems can be attacked in new ways * AI embeds values about what is human * AI tells us about our own intelligence