Announcements Homework 6 due tonight @ 11:59pm AI seminar today (Speech): Speaker localization using CNNs SEIs active now.

Slides:



Advertisements
Similar presentations
Review 23 April. Group exercise  What lessons have you learned?  Write full sentences until I tell you to stop.  Examples: Code can affect behavior.
Advertisements

13 Chapter Marketing in Today’s World pp
These are two population pyramids, they show the numbers of males and females and their ages in a given country. These pyramids represent Russia and Ghana.
Jason Vander Weele, Analyst Lakeshore Technical College April 24, 2014 Madison College IR State-Called Meeting.
303KM Project Management1 Chapter 2: The Project Management in Context of Organization Environment.
Research Methods in Psychology (Pp ). IB Internal Assessment The IB Psychology Guide states that SL students are required to replicate a simple.
Crowdsourced Enumeration Queries Ruihan Shan. Introduction Motivation.
Section 1. Qualitative Research: Theory and Practice  Methods chosen for research dependant on a number of factors including:  Purpose of the research.
Can We Trust the Computer? FIRE, Chapter 4. What Can Go Wrong? What are the risks and reasons for computer failures? How much risk must or should we accept?
Computational Models of Discourse Analysis Carolyn Penstein Rosé Language Technologies Institute/ Human-Computer Interaction Institute.
WoSA Assessment Training Module 2: Basic Principles of Needs Assessments.
Lecture Notes and Electronic Presentations, © 2013 Dr. Kelly Significance and Sample Size Refresher Harrison W. Kelly III, Ph.D. Lecture # 3.
Surf smart training.
Citizen Journalism and Ethics
AP Biology Discussion Notes
Top 20 Career Attraction Tips
Deliver On-the-Job Training
Business Models, Revenue Models & the Lean Startup Methodology
AP CSP: What is Big Data?.
Fundamentals of Computer Systems
Evaluating Classifiers
Highlighting a Module 2 Lesson: Secondary
Section 2: Statistics and Models
AF1: Thinking Scientifically
Panagiotis Demestichas
Breaking Down Essay Prompts
Section 2: Statistics and Models
Tuesday August 23,2016 Notes –Binder Check - 08/14, every work should be completed. GPS – SEV5. Students will recognize that human beings are part of the.
CHAPTER 10 Estimating with Confidence
Chapter 2: The Project Management and Information Technology Context
CHAPTER 8 Estimating with Confidence
Check Your Assumptions
Investigations [ 10 week ] Investigations 2015 second semester.
Announcements Homework 6 due Friday 11:59pm.
Correlation and Regression
Introductory Statistical Language
How to evaluate a piece of research
PAF 101 Module 5, Lecture 5 “Snowflakes are one of nature’s most fragile things, but just look at what they can do when they stick together.” 1.
THE NATURE of LEARNER LANGUAGE
ETHICS BOWL kantian ETHICS.
I271B Quantitative Methods
Recurrent Neural Networks
ELEC4011 Ethics & Electrical Engineering Practice Hugh Outhred
The Scientific Method.
What is qualitative research?
Section 2: Statistics and Models
Word Embedding Word2Vec.
Ethical, Legal, Cultural and Environmental Concerns
THE NEW WORLD OF POLICE ACCOUNTABILITY
Setting Healthy Eating & Physical Activity Goals
The Nature of Learner Language (Chapter 2 Rod Ellis, 1997) Page 15
Human Resource Management and Labor Relations
Ethical Hacking.
Business Ethics and Social Responsibility
Learning Targets Students will be able to: Compare linear, quadratic, and exponential models and given a set of data, decide which type of function models.
Chapter 13 Marketing in Today’s World
Chapter 8: Estimating With Confidence
Legal Issues in Podcasting: What Broadcasters Need to Know
Quantitative Research
Ecolog.
Coming up soon Monday: Searching the Literature Quiz
Process of the Scientific Method
Software Requirements
Mikael Olsson Control Engineer
Evaluation: Inspections, Analytics & Models
Matthew Gerrick Tyleik McLaughlin Jaylen Killens 2nd period
Instructor: Vincent Conitzer
Important Questions.
WSP™ Inventory Equipping Certification – Focus – What You Need to Know
Evaluation David Kauchak CS 158 – Fall 2019.
Presentation transcript:

Announcements Homework 6 due tonight @ 11:59pm AI seminar today (Speech): Speaker localization using CNNs SEIs active now

Today’s learning goals At the end of today, you should be able to Describe how ethical concerns fit into data collection, model design, and applications Explain where recent ethical failures in AI went wrong

The way it’s viewed now

Facebook’s response Chief Security Officer at Facebook

Thinking intelligently about AI ethics Aside from general ethical issues, we can break AI-specific ethical issues down into three areas: Data collection – is your process of collecting data ethical? Model design – are the assumptions you make in your model ethical? Model application – is the way you use your model ethical?

Data collection Engineering task – collect enough data to train the model From a population that represents the target Important questions Have you captured the diversity of the real target population? Collecting political speech on Twitter; just collecting liberal speech biases the model! Are you collecting more data than you need/should? Never need SSNs, for example Should you use income level in predicting criminal risk?

Model design Engineering task – build the model structure around assumptions about the data distribution Important questions Could those assumptions be harmful to users? Using an offline model assumes that test points are independent But what if the order of test points should actually affect the model? (e.g., “I’ll kill him when I see him”; need context to say if “kill” is literal or figurative) Are the features you’re using biased? Number of traffic stops by police

Model application Engineering task – use the trained model to make predictions on data from the right kind of distribution Important questions Is the purpose that you’re using this for ethical? Applying a sentiment classification model to find negative speech about a political leader Is the way that you’re getting/processing predictions ethical? Air traffic control system that assumes its predictor of runway status is always accurate

Exercise: Analyzing ethical breaches

Next time Review session for final: BRING QUESTIONS!