Sergey Karayev, profile picture
Uploaded bySergey Karayev
PDF, PPTX10,504 views

Lecture 9: AI Ethics (Full Stack Deep Learning - Spring 2021)

This document discusses ethics in artificial intelligence and machine learning. It begins with an outline of the topics to be covered, including what ethics is, long-term and near-term ethical problems in AI, best practices, and where to learn more. It then covers the definitions of different ethical theories, uses "trolley problems" to illustrate ethical intuitions, and notes there is no clear consensus among ethical theories. The document discusses long-term problems like autonomous weapons, AI replacing human labor, and ensuring AI is aligned with human values and goals. Near-term problems discussed include bias in hiring models, fairness in models like COMPAS, and issues of representation in data and AI systems.

Technology◦
Related topics:
Deep Learning•

Embed presentation

Download as PDF, PPTX
Full Stack Deep Learning - UC Berkeley Spring 2021 - Sergey Karayev, Josh Tobin, Pieter Abbeel Ethics
Full Stack Deep Learning - UC Berkeley Spring 2021 Preamble • This is a huge subject, spanning many disciplines and addressing many real different problems. • We, ML practitioners, need to have a student mindset, and not assume we have the answers. These are not easy problems. • I am not an expert. Excellent resources recommended at the end. 2
Full Stack Deep Learning - UC Berkeley Spring 2021 Tech Ethics Curricula • 115 university tech ethics courses analyzed • Huge variation in materials • More consensus in outcomes: ability to critique, spot issues, and make and communicate arguments. 3 SIGCSE 2020
Full Stack Deep Learning - UC Berkeley Spring 2021 One last preamble 4 https://www.newyorker.com/books/page-turner/this-is-water
Full Stack Deep Learning - UC Berkeley Spring 2021 Outline • What is ethics • Long-term ethical problems in AI • Near-term ethical problems in AI • Best practices • Where to learn more 5
Full Stack Deep Learning - UC Berkeley Spring 2021 What is ethics 6
Full Stack Deep Learning - UC Berkeley Spring 2021 Ethics? 7 Ethics ≠ Feelings Ethics ≠ Laws Ethics ≠ Societal Beliefs https://www.scu.edu/ethics/ethics-resources/ethical-decision-making/what-is-ethics/
Full Stack Deep Learning - UC Berkeley Spring 2021 Ethical Theories 8 https://kevinbinz.com/2017/04/13/ethical-theory-intro/ Divine Command - Moral behaviors are those commanded by the divine - Criticism: not much philosophy can say Virtue Ethics - Moral behaviors uphold the person's virtues - Criticism: increasing evidence that character traits are illusory Deontology (Duty) - Moral behaviors are those that satisfy the categorical imperative (e.g. don't lie, don't kill) - Criticism: unacceptable inflexibility Utilitarianism - Moral behaviors are those that bring the most good to the most people - Criticism: How to measure utility?
Full Stack Deep Learning - UC Berkeley Spring 2021 Is one a clear winner? • Professional philosophers are just about evenly split between these 9 https://philpapers.org/surveys/results.pl
Full Stack Deep Learning - UC Berkeley Spring 2021 Intuition through "Trolley Problems" 10
Full Stack Deep Learning - UC Berkeley Spring 2021 Trolley Problems 11
Full Stack Deep Learning - UC Berkeley Spring 2021 Ethical Theories 12 https://www.researchgate.net/figure/Summary-of-major-ethical-theories_tbl2_229658531
Full Stack Deep Learning - UC Berkeley Spring 2021 Ethics of Technology • Ethics change with technological progress • e.g. Industrial revolution • e.g. Right to Internet access • e.g. Birth control, surrogate pregnancy, embryo selection, artificial womb • e.g. Lab-grown meat 13
Full Stack Deep Learning - UC Berkeley Spring 2021 Questions? 14
Full Stack Deep Learning - UC Berkeley Spring 2021 Long-term problems 15
Full Stack Deep Learning - UC Berkeley Spring 2021 Autonomous Weapons • Tempting to dismiss as far-fetched at this time • But "the future is already here, just not evenly distributed" 16
Full Stack Deep Learning - UC Berkeley Spring 2021 Replacing human labor • AI and robots replacing humans in existing jobs • Good and bad! • Interesting spin: AI controlling human labor 17 Sfdpnnfoefe! tipsu!tupsz
Full Stack Deep Learning - UC Berkeley Spring 2021 Replacing humans entirely 18
Full Stack Deep Learning - UC Berkeley Spring 2021 What's common in all these long-term problems? 19
Full Stack Deep Learning - UC Berkeley Spring 2021 Alignment • "Paperclip maximizer": AGI given the goal of producing paperclips eventually turns every atom in space into paperclips • Old lesson: establishing and communicating our goals and values is hard, and technology amplifies the difficulty 20
Full Stack Deep Learning - UC Berkeley Spring 2021 Guiding principle: AI systems we build need to be aligned with our goals and values 21
Full Stack Deep Learning - UC Berkeley Spring 2021 Questions? 22
Full Stack Deep Learning - UC Berkeley Spring 2021 Deep topic • Active area of research (active at Berkeley) • Useful lens for near-term problems, too 23
Full Stack Deep Learning - UC Berkeley Spring 2021 Near-term problems 24
Full Stack Deep Learning - UC Berkeley Spring 2021 Hiring 25
Full Stack Deep Learning - UC Berkeley Spring 2021 26 https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias- against-women-idUSKCN1MK08G
Full Stack Deep Learning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 27 ML Model
Full Stack Deep Learning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 28 ML Model Data Ijsjoh!efdjtjpot@! Ps!kpc!qfsgpsnbodf@
Full Stack Deep Learning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 29 ML Model Data World Lopxo!up!cf!cjbtfe!jo!nboz!xbzt;! .!ijsjoh!qjqfmjof! .!ijsjoh!efdjtjpot! .!qfsgpsnbodf!sbujoht Tp!op!nbuufs!xibu!xf!qjdl-! pvs!ebub!jt!bmtp!cjbtfe Uifsfgpsf!pvs! npefm!jt!cjbtfe
Full Stack Deep Learning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 30 Action Tpvsdjoh@!Epvcmf.difdljoh!ivnbo! efdjtjpo@!Ijsjoh@ ML Model Data World
Full Stack Deep Learning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 31 Action Jo!nboz!dbtft-!bnqmjgzjoh! fyjtujoh!cjbtft" ML Model Data World
Full Stack Deep Learning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume • Amplifying existing biases is not aligned with our goals and values! 32 Action Jo!nboz!dbtft-!bnqmjgzjoh! fyjtujoh!cjbtft" ML Model Data World
Full Stack Deep Learning - UC Berkeley Spring 2021 Questions? 33
Full Stack Deep Learning - UC Berkeley Spring 2021 Fairness 34
Full Stack Deep Learning - UC Berkeley Spring 2021 COMPAS • Goal: predict recidivism, such that judges can consult 1-10 score in pre-trial sentencing decisions. • Motivation: be less biased than humans • Solution: • Gather data • Exclude protected class attributes (race, etc) • Ensure that our model's score corresponds to same probability of recidivism across all groups 35 Correctional Offender Management Profiling for Alternative Sanctions
Full Stack Deep Learning - UC Berkeley Spring 2021 And yet! 36 https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing
Full Stack Deep Learning - UC Berkeley Spring 2021 21 Fairness Definitions 37 Aravind Narayanan - 21 Fairness Definitions 🙏
Full Stack Deep Learning - UC Berkeley Spring 2021 Statistical Bias • Statistical bias: Difference between estimator's expected value and the true value • In this sense, COMPAS scores are not biased, w.r.t re-arrest • But is it an adequate fairness criterion? Is it aligned with our values? 38 https://www.washingtonpost.com/news/monkey-cage/wp/2016/10/17/can-an-algorithm-be-racist-our-analysis-is-more-cautious- than-propublicas/ Jnqpsubou!dbwfbu"!Xf!pomz! ibwf!ebub!gps!bssftut-!opu! dsjnft!dpnnjuufe/!Uifsf!nbz! xfmm!cf!cjbt!jo!bssftut/
Full Stack Deep Learning - UC Berkeley Spring 2021 Different perspectives on fairness • What do different stakeholders want from the classifier? • Decision-maker: of those I've labeled high risk, how many recidivated? • Predictive value: TP / (TP + FP) • Defendant: what's the probability I'll be incorrectly classified as high risk? • False positive rate: FP / (FP + TN) • Society: is the selected set demographically balanced? • Demographic parity 39 Aravind Narayanan - 21 Fairness Definitions
Full Stack Deep Learning - UC Berkeley Spring 2021 Group Fairness • Do outcomes differ between groups (e.g. demographic), which we have no reason to believe are actually different? • Motivation of the Pro-Publica article 40 Aravind Narayanan - 21 Fairness Definitions Gbmtf! Qptjujwf Gbmtf! Ofhbujwf
Full Stack Deep Learning - UC Berkeley Spring 2021 No "correct" group fairness definition 41 Aravind Narayanan - 21 Fairness Definitions
Full Stack Deep Learning - UC Berkeley Spring 2021 It gets worse... • Okay, then let's just pick most important two metrics (FPR and FNR), and allow the model to use protected class attributes. • Now we fail individual fairness -> • Fine, let's just pick one! (e.g. equal FPR) • We still sacrifice some utility (e.g. public safety, or number of defendants released) 42 Aravind Narayanan - 21 Fairness Definitions
Full Stack Deep Learning - UC Berkeley Spring 2021 Interactive Example 43 https://research.google.com/bigpicture/attacking-discrimination-in-ml/
Full Stack Deep Learning - UC Berkeley Spring 2021 By the way... 44 Aravind Narayanan - 21 Fairness Definitions
Full Stack Deep Learning - UC Berkeley Spring 2021 Trade-offs 45 Aravind Narayanan - 21 Fairness Definitions
Full Stack Deep Learning - UC Berkeley Spring 2021 Seeing the water 46
Full Stack Deep Learning - UC Berkeley Spring 2021 47 https://www.mobilizegreen.org/blog/2018/9/30/environmental-equity-vs-environmental-justice-whats-the-difference
Full Stack Deep Learning - UC Berkeley Spring 2021 48 https://www.mobilizegreen.org/blog/2018/9/30/environmental-equity-vs-environmental-justice-whats-the-difference
Full Stack Deep Learning - UC Berkeley Spring 2021 49 https://www.mobilizegreen.org/blog/2018/9/30/environmental-equity-vs-environmental-justice-whats-the-difference
Full Stack Deep Learning - UC Berkeley Spring 2021 Questions? 50
Full Stack Deep Learning - UC Berkeley Spring 2021 Representation 51
Full Stack Deep Learning - UC Berkeley Spring 2021 Representation 52 https://twitter.com/nke_ise/status/897756900753891328
Full Stack Deep Learning - UC Berkeley Spring 2021 Not a new problem, sadly 53
Full Stack Deep Learning - UC Berkeley Spring 2021 ...or a problem in just our field 54 Recent improvement!
Full Stack Deep Learning - UC Berkeley Spring 2021 Large part of the solution 55 https://www.nytimes.com/2021/03/15/technology/artificial-intelligence-google-bias.html
Full Stack Deep Learning - UC Berkeley Spring 2021 Another example 56 Recent improvement!
Full Stack Deep Learning - UC Berkeley Spring 2021 Translation 57 Aravind Narayanan - 21 Fairness Definitions Recent improvement!
Full Stack Deep Learning - UC Berkeley Spring 2021 Word Embeddings • Word2Vec introduced vector math on word embeddings • Reveal harmful biases encoded in our language corpora • Potential solution: de-bias at training time, but at least make user aware 58
Full Stack Deep Learning - UC Berkeley Spring 2021 GPT-3 59
Full Stack Deep Learning - UC Berkeley Spring 2021 Recent Language Models • Paper apparently led to authors being fired by Google 60 http://faculty.washington.edu/ebender/papers/Stochastic_Parrots.pdf
Full Stack Deep Learning - UC Berkeley Spring 2021 Seeing the water • Whether language models should reflect the world as it is in the data, or as society believes it should be, depends on what use they are applied to. • And when we do want them to reflect the world as it should be, do we agree on what that is? 61
Full Stack Deep Learning - UC Berkeley Spring 2021 Face Recognition • Old context: no expectation of privacy in public • New context: "in public" now means "on any street in any city, or on any website on the internet" • Is it ethical to work on this? • Is it a problem if it does not work as well on some ethnicities? 62 https://fairmlbook.org/introduction.html
Full Stack Deep Learning - UC Berkeley Spring 2021 Questions? 63
Full Stack Deep Learning - UC Berkeley Spring 2021 Best practices 64
Full Stack Deep Learning - UC Berkeley Spring 2021 What do practitioners need? 65 • support in fairness-aware data collection and curation • overcoming teams’ blind spots • implementing more proactive fairness auditing processes • auditing complex ML systems • deciding how to address particular instances of unfairness • addressing biases in the humans embedded throughout the ML development pipeline 2019
Full Stack Deep Learning - UC Berkeley Spring 2021 Some suggestions • Ethical risk sweeping
 treat like cybersecurity penetration testing • Expanding the ethical circle
 whose interests, desires, experiences, values have we just assumed instead of consulted? • Think about the terrible people
 Who might abuse, steal, weaponize what we build? What incentives are we creating? • Closing the loop
 Remember that this is not a process to complete and forget. Set up ways to keep improving. 66 https://www.youtube.com/watch?v=av7utkFXbU4
Full Stack Deep Learning - UC Berkeley Spring 2021 Model Cards 67 https://modelcards.withgoogle.com/face-detection
Full Stack Deep Learning - UC Berkeley Spring 2021 Bias Audit 68 http://www.datasciencepublicpolicy.org/projects/aequitas/
Full Stack Deep Learning - UC Berkeley Spring 2021 69 http://www.datasciencepublicpolicy.org/projects/aequitas/
Full Stack Deep Learning - UC Berkeley Spring 2021 70 AI bias is rooted in societal inequities, manifested in data, and amplified by AI. Unbiased societal AI is an impossible goal for a single project to achieve, but greatly mitigating its harmful effects is not. Visualization courtesy of Eric Wang
Full Stack Deep Learning - UC Berkeley Spring 2021 71 Build a diverse AI team to bring lived experience to their work Display insights about bias contextually and intuitively in product Make AI fairness and transparency a key component of our work Visualization courtesy of Eric Wang
Full Stack Deep Learning - UC Berkeley Spring 2021 A professional code of ethics? 72
Full Stack Deep Learning - UC Berkeley Spring 2021 Medicine 73
Full Stack Deep Learning - UC Berkeley Spring 2021 Armed Forces 74
Full Stack Deep Learning - UC Berkeley Spring 2021 75
Full Stack Deep Learning - UC Berkeley Spring 2021 Questions? 76
Full Stack Deep Learning - UC Berkeley Spring 2021 Learn more 77
Full Stack Deep Learning - UC Berkeley Spring 2021 Experts • Practical Data Ethics course from Rachel Thomas: https://ethics.fast.ai • Single lecture from Fast.ai 2020 course: https:// www.youtube.com/watch?v=krIVOb23EH8 78
Full Stack Deep Learning - UC Berkeley Spring 2021 79 https://fairmlbook.org
Full Stack Deep Learning - UC Berkeley Spring 2021 80 https://dssg.github.io/fairness_tutorial/
Full Stack Deep Learning - UC Berkeley Spring 2021 81
Full Stack Deep Learning - UC Berkeley Spring 2021 82
Full Stack Deep Learning - UC Berkeley Spring 2021 Thank you! 83

More Related Content

PDF
Lecture 1: Deep Learning Fundamentals - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 
PDF
Lecture 10: ML Testing & Explainability (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PDF
Lecture 5: ML Projects (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PDF
Lecture 8: Data Management (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PDF
Lecture 2.B: Computer Vision Applications - Full Stack Deep Learning - Spring...
bySergey Karayev
 
PDF
Lecture 7: Troubleshooting Deep Neural Networks (Full Stack Deep Learning - S...
bySergey Karayev
 
PDF
Lecture 13: ML Teams (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PDF
Troubleshooting Deep Neural Networks - Full Stack Deep Learning
bySergey Karayev
 
Lecture 1: Deep Learning Fundamentals - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 
Lecture 10: ML Testing & Explainability (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
Lecture 5: ML Projects (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
Lecture 8: Data Management (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
Lecture 2.B: Computer Vision Applications - Full Stack Deep Learning - Spring...
bySergey Karayev
 
Lecture 7: Troubleshooting Deep Neural Networks (Full Stack Deep Learning - S...
bySergey Karayev
 
Lecture 13: ML Teams (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
Troubleshooting Deep Neural Networks - Full Stack Deep Learning
bySergey Karayev
 

What's hot

PDF
Lecture 4: Transformers (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PDF
On Impact in Software Engineering Research (Dagstuhl 2020)
byCISPA Helmholtz Center for Information Security
 
PDF
Deep learning with Keras
byQuantUniversity
 
PDF
Lecture 2.A: Convolutional Networks - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 
PDF
Data Management - Full Stack Deep Learning
bySergey Karayev
 
PDF
Test-Driven Machine Learning
byC4Media
 
PDF
Lab 1: Intro and Setup - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 
PDF
On Impact in Software Engineering Research (HU Berlin 2021)
byCISPA Helmholtz Center for Information Security
 
PDF
Testing and Deployment - Full Stack Deep Learning
bySergey Karayev
 
PDF
On Impact in Software Engineering Research
byCISPA Helmholtz Center for Information Security
 
PDF
On impact in Software Engineering Research (ICSE 2018 New Faculty Symposium)
byCISPA Helmholtz Center for Information Security
 
PDF
Enabling Congestion Control Using Homogeneous Archetypes
byJames Johnson
 
PDF
Lecture 3: RNNs - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 
Lecture 4: Transformers (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
On Impact in Software Engineering Research (Dagstuhl 2020)
byCISPA Helmholtz Center for Information Security
 
Deep learning with Keras
byQuantUniversity
 
Lecture 2.A: Convolutional Networks - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 
Data Management - Full Stack Deep Learning
bySergey Karayev
 
Test-Driven Machine Learning
byC4Media
 
Lab 1: Intro and Setup - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 
On Impact in Software Engineering Research (HU Berlin 2021)
byCISPA Helmholtz Center for Information Security
 
Testing and Deployment - Full Stack Deep Learning
bySergey Karayev
 
On Impact in Software Engineering Research
byCISPA Helmholtz Center for Information Security
 
On impact in Software Engineering Research (ICSE 2018 New Faculty Symposium)
byCISPA Helmholtz Center for Information Security
 
Enabling Congestion Control Using Homogeneous Archetypes
byJames Johnson
 
Lecture 3: RNNs - Full Stack Deep Learning - Spring 2021
bySergey Karayev
 

Similar to Lecture 9: AI Ethics (Full Stack Deep Learning - Spring 2021)

PDF
M2 l10 fairness, accountability, and transparency
byBoPeng76
 
PDF
Introduction to AI Ethics
byGabriele Graffieti
 
PPTX
Introduction to Deep Learning and ML.pptx
bysilaya6647
 
PPTX
Introduction to Deep Learning and ML.pptx
bysilaya6647
 
PDF
Lecture 11: ML Deployment & Monitoring (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PPTX
Responsible AI in Industry (Tutorials at AAAI 2021, FAccT 2021, and WWW 2021)
byKrishnaram Kenthapadi
 
PPT
PROFESSIONAL ELECTIVE - CCS345 - Ethics & AI unit 2.ppt
bycsedeptsrpce
 
PDF
Beyond-Accuracy Perspectives: Explainability and Fairness
byErasmo Purificato
 
PDF
Engineering Ethics: Practicing Fairness
byClare Corthell
 
PPTX
Responsible AI in Industry (ICML 2021 Tutorial)
byKrishnaram Kenthapadi
 
PDF
Fair AI
byVidhya Chandrasekaran
 
PPTX
Ethical_and_Social_Implications_AI_ML_pptx
bynbtalk03
 
PPTX
Session 2.pptxqfqfefefqefqefqefqefqfqefqefqef
byKaranveerBhatt
 
PPTX
Fairness and Privacy in AI/ML Systems
byKrishnaram Kenthapadi
 
PPTX
Ethical Issues in Machine Learning Algorithms. (Part 1)
byVladimir Kanchev
 
PPTX
Responsible AI in Industry: Practical Challenges and Lessons Learned
byKrishnaram Kenthapadi
 
PDF
Regoli fairness deep_learningitalia_20220127
byMatteo Testi
 
PPTX
Session 5 coding handson Tensorflow
byRajagopal A
 
PPTX
Ramon van den Akker. Fairness of machine learning models an overview and prac...
byLviv Startup Club
 
PPTX
Dev fest 2020 norway
bySharmistha Chatterjee
 
M2 l10 fairness, accountability, and transparency
byBoPeng76
 
Introduction to AI Ethics
byGabriele Graffieti
 
Introduction to Deep Learning and ML.pptx
bysilaya6647
 
Introduction to Deep Learning and ML.pptx
bysilaya6647
 
Lecture 11: ML Deployment & Monitoring (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
Responsible AI in Industry (Tutorials at AAAI 2021, FAccT 2021, and WWW 2021)
byKrishnaram Kenthapadi
 
PROFESSIONAL ELECTIVE - CCS345 - Ethics & AI unit 2.ppt
bycsedeptsrpce
 
Beyond-Accuracy Perspectives: Explainability and Fairness
byErasmo Purificato
 
Engineering Ethics: Practicing Fairness
byClare Corthell
 
Responsible AI in Industry (ICML 2021 Tutorial)
byKrishnaram Kenthapadi
 
Fair AI
byVidhya Chandrasekaran
 
Ethical_and_Social_Implications_AI_ML_pptx
bynbtalk03
 
Session 2.pptxqfqfefefqefqefqefqefqfqefqefqef
byKaranveerBhatt
 
Fairness and Privacy in AI/ML Systems
byKrishnaram Kenthapadi
 
Ethical Issues in Machine Learning Algorithms. (Part 1)
byVladimir Kanchev
 
Responsible AI in Industry: Practical Challenges and Lessons Learned
byKrishnaram Kenthapadi
 
Regoli fairness deep_learningitalia_20220127
byMatteo Testi
 
Session 5 coding handson Tensorflow
byRajagopal A
 
Ramon van den Akker. Fairness of machine learning models an overview and prac...
byLviv Startup Club
 
Dev fest 2020 norway
bySharmistha Chatterjee
 

More from Sergey Karayev

PDF
Lecture 6: Infrastructure & Tooling (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PDF
AI Masterclass at ASU GSV 2019
bySergey Karayev
 
PDF
Machine Learning Teams - Full Stack Deep Learning
bySergey Karayev
 
PDF
Setting up Machine Learning Projects - Full Stack Deep Learning
bySergey Karayev
 
KEY
Attentional Object Detection - introductory slides.
bySergey Karayev
 
PDF
Lecture 12: Research Directions (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
PDF
Research Directions - Full Stack Deep Learning
bySergey Karayev
 
PDF
Infrastructure and Tooling - Full Stack Deep Learning
bySergey Karayev
 
Lecture 6: Infrastructure & Tooling (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
AI Masterclass at ASU GSV 2019
bySergey Karayev
 
Machine Learning Teams - Full Stack Deep Learning
bySergey Karayev
 
Setting up Machine Learning Projects - Full Stack Deep Learning
bySergey Karayev
 
Attentional Object Detection - introductory slides.
bySergey Karayev
 
Lecture 12: Research Directions (Full Stack Deep Learning - Spring 2021)
bySergey Karayev
 
Research Directions - Full Stack Deep Learning
bySergey Karayev
 
Infrastructure and Tooling - Full Stack Deep Learning
bySergey Karayev
 

Recently uploaded

PDF
Generating Structured Outputs from Unstructured Content Using LLMs
byEnterprise Knowledge
 
PDF
DataLiva-LivaGRC- Business Software for Governance, Risk, and Compliance
byMustafa Kuğu
 
PDF
Building Software in the AI Era: Spec-Driven Development with Kiro IDE
byHaim Michael
 
PDF
splunk-peak-threat-hunting-framework.pdf
bylobster6719
 
PDF
Azure Identity, Governance, and Monitoring solutions
byVICTOR MAESTRE RAMIREZ
 
PDF
Top 10 API Automation Testing Tools: Features, Pros & Cons
byhenrycavill30521
 
PPTX
Large Language Model. LLM Audit Artificial Intelligence
byMANASCHOUDHARY22
 
PDF
oracle_apex_tamil_export_an_application.pdf
byUthamaselvan M
 
PDF
Chapter 1 Introduction to Computer Security.pdf
byGetnet Tigabie Askale -(GM)
 
PDF
How Application Performance Monitoring Tools Are Used in Performance Testing
byhenrycavill30521
 
PDF
What is Voice User Interface (VUI) Definition & Examples.pdf
byDesign Studio UI UX
 
PDF
The_Boardroom_Cyber_Playbook_Research_Edition
bySaraDavis91
 
PDF
Benefits of Using the IAC 500 Sensor for Ambient Conditions
byCS Instruments
 
PDF
Beyond BBB: Practical Alternatives to Posterior Approximation in Bayesian Neu...
byTomasz Kusmierczyk
 
PPTX
IBM_NEXA_DAC2021 NEXA, a cloud-native platform for collaborative hardware log...
byArun Joseph
 
PPTX
Why 2026 Could Be a Turning Point for Decentralized Exchanges.pptx
bycalliemorgan193
 
PPTX
AI Bot Traffic Surge: Retail Fraud Threat for Age-Restricted Websites
byFTx Identity
 
PPTX
Sensors-and-Actuators-in-IoT-Systems.pptx
byMuhammedNihal54
 
PDF
Achieving Interoperability in Healthcare IT: A Strategic Guide
byHart, Inc.
 
PDF
Scott M. Graffius' Phases of Team Development - Applied to Human Teams and Hu...
byScott M. Graffius
 
Generating Structured Outputs from Unstructured Content Using LLMs
byEnterprise Knowledge
 
DataLiva-LivaGRC- Business Software for Governance, Risk, and Compliance
byMustafa Kuğu
 
Building Software in the AI Era: Spec-Driven Development with Kiro IDE
byHaim Michael
 
splunk-peak-threat-hunting-framework.pdf
bylobster6719
 
Azure Identity, Governance, and Monitoring solutions
byVICTOR MAESTRE RAMIREZ
 
Top 10 API Automation Testing Tools: Features, Pros & Cons
byhenrycavill30521
 
Large Language Model. LLM Audit Artificial Intelligence
byMANASCHOUDHARY22
 
oracle_apex_tamil_export_an_application.pdf
byUthamaselvan M
 
Chapter 1 Introduction to Computer Security.pdf
byGetnet Tigabie Askale -(GM)
 
How Application Performance Monitoring Tools Are Used in Performance Testing
byhenrycavill30521
 
What is Voice User Interface (VUI) Definition & Examples.pdf
byDesign Studio UI UX
 
The_Boardroom_Cyber_Playbook_Research_Edition
bySaraDavis91
 
Benefits of Using the IAC 500 Sensor for Ambient Conditions
byCS Instruments
 
Beyond BBB: Practical Alternatives to Posterior Approximation in Bayesian Neu...
byTomasz Kusmierczyk
 
IBM_NEXA_DAC2021 NEXA, a cloud-native platform for collaborative hardware log...
byArun Joseph
 
Why 2026 Could Be a Turning Point for Decentralized Exchanges.pptx
bycalliemorgan193
 
AI Bot Traffic Surge: Retail Fraud Threat for Age-Restricted Websites
byFTx Identity
 
Sensors-and-Actuators-in-IoT-Systems.pptx
byMuhammedNihal54
 
Achieving Interoperability in Healthcare IT: A Strategic Guide
byHart, Inc.
 
Scott M. Graffius' Phases of Team Development - Applied to Human Teams and Hu...
byScott M. Graffius
 

Lecture 9: AI Ethics (Full Stack Deep Learning - Spring 2021)

  • 1.
    Full Stack DeepLearning - UC Berkeley Spring 2021 - Sergey Karayev, Josh Tobin, Pieter Abbeel Ethics
  • 2.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Preamble • This is a huge subject, spanning many disciplines and addressing many real different problems. • We, ML practitioners, need to have a student mindset, and not assume we have the answers. These are not easy problems. • I am not an expert. Excellent resources recommended at the end. 2
  • 3.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Tech Ethics Curricula • 115 university tech ethics courses analyzed • Huge variation in materials • More consensus in outcomes: ability to critique, spot issues, and make and communicate arguments. 3 SIGCSE 2020
  • 4.
    Full Stack DeepLearning - UC Berkeley Spring 2021 One last preamble 4 https://www.newyorker.com/books/page-turner/this-is-water
  • 5.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Outline • What is ethics • Long-term ethical problems in AI • Near-term ethical problems in AI • Best practices • Where to learn more 5
  • 6.
    Full Stack DeepLearning - UC Berkeley Spring 2021 What is ethics 6
  • 7.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Ethics? 7 Ethics ≠ Feelings Ethics ≠ Laws Ethics ≠ Societal Beliefs https://www.scu.edu/ethics/ethics-resources/ethical-decision-making/what-is-ethics/
  • 8.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Ethical Theories 8 https://kevinbinz.com/2017/04/13/ethical-theory-intro/ Divine Command - Moral behaviors are those commanded by the divine - Criticism: not much philosophy can say Virtue Ethics - Moral behaviors uphold the person's virtues - Criticism: increasing evidence that character traits are illusory Deontology (Duty) - Moral behaviors are those that satisfy the categorical imperative (e.g. don't lie, don't kill) - Criticism: unacceptable inflexibility Utilitarianism - Moral behaviors are those that bring the most good to the most people - Criticism: How to measure utility?
  • 9.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Is one a clear winner? • Professional philosophers are just about evenly split between these 9 https://philpapers.org/surveys/results.pl
  • 10.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Intuition through "Trolley Problems" 10
  • 11.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Trolley Problems 11
  • 12.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Ethical Theories 12 https://www.researchgate.net/figure/Summary-of-major-ethical-theories_tbl2_229658531
  • 13.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Ethics of Technology • Ethics change with technological progress • e.g. Industrial revolution • e.g. Right to Internet access • e.g. Birth control, surrogate pregnancy, embryo selection, artificial womb • e.g. Lab-grown meat 13
  • 14.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Questions? 14
  • 15.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Long-term problems 15
  • 16.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Autonomous Weapons • Tempting to dismiss as far-fetched at this time • But "the future is already here, just not evenly distributed" 16
  • 17.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Replacing human labor • AI and robots replacing humans in existing jobs • Good and bad! • Interesting spin: AI controlling human labor 17 Sfdpnnfoefe! tipsu!tupsz
  • 18.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Replacing humans entirely 18
  • 19.
    Full Stack DeepLearning - UC Berkeley Spring 2021 What's common in all these long-term problems? 19
  • 20.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Alignment • "Paperclip maximizer": AGI given the goal of producing paperclips eventually turns every atom in space into paperclips • Old lesson: establishing and communicating our goals and values is hard, and technology amplifies the difficulty 20
  • 21.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Guiding principle: AI systems we build need to be aligned with our goals and values 21
  • 22.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Questions? 22
  • 23.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Deep topic • Active area of research (active at Berkeley) • Useful lens for near-term problems, too 23
  • 24.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Near-term problems 24
  • 25.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Hiring 25
  • 26.
    Full Stack DeepLearning - UC Berkeley Spring 2021 26 https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias- against-women-idUSKCN1MK08G
  • 27.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 27 ML Model
  • 28.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 28 ML Model Data Ijsjoh!efdjtjpot@! Ps!kpc!qfsgpsnbodf@
  • 29.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 29 ML Model Data World Lopxo!up!cf!cjbtfe!jo!nboz!xbzt;! .!ijsjoh!qjqfmjof! .!ijsjoh!efdjtjpot! .!qfsgpsnbodf!sbujoht Tp!op!nbuufs!xibu!xf!qjdl-! pvs!ebub!jt!bmtp!cjbtfe Uifsfgpsf!pvs! npefm!jt!cjbtfe
  • 30.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 30 Action Tpvsdjoh@!Epvcmf.difdljoh!ivnbo! efdjtjpo@!Ijsjoh@ ML Model Data World
  • 31.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume 31 Action Jo!nboz!dbtft-!bnqmjgzjoh! fyjtujoh!cjbtft" ML Model Data World
  • 32.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Hiring • ML model to predict hiring decision given a resume • Amplifying existing biases is not aligned with our goals and values! 32 Action Jo!nboz!dbtft-!bnqmjgzjoh! fyjtujoh!cjbtft" ML Model Data World
  • 33.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Questions? 33
  • 34.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Fairness 34
  • 35.
    Full Stack DeepLearning - UC Berkeley Spring 2021 COMPAS • Goal: predict recidivism, such that judges can consult 1-10 score in pre-trial sentencing decisions. • Motivation: be less biased than humans • Solution: • Gather data • Exclude protected class attributes (race, etc) • Ensure that our model's score corresponds to same probability of recidivism across all groups 35 Correctional Offender Management Profiling for Alternative Sanctions
  • 36.
    Full Stack DeepLearning - UC Berkeley Spring 2021 And yet! 36 https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing
  • 37.
    Full Stack DeepLearning - UC Berkeley Spring 2021 21 Fairness Definitions 37 Aravind Narayanan - 21 Fairness Definitions 🙏
  • 38.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Statistical Bias • Statistical bias: Difference between estimator's expected value and the true value • In this sense, COMPAS scores are not biased, w.r.t re-arrest • But is it an adequate fairness criterion? Is it aligned with our values? 38 https://www.washingtonpost.com/news/monkey-cage/wp/2016/10/17/can-an-algorithm-be-racist-our-analysis-is-more-cautious- than-propublicas/ Jnqpsubou!dbwfbu"!Xf!pomz! ibwf!ebub!gps!bssftut-!opu! dsjnft!dpnnjuufe/!Uifsf!nbz! xfmm!cf!cjbt!jo!bssftut/
  • 39.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Different perspectives on fairness • What do different stakeholders want from the classifier? • Decision-maker: of those I've labeled high risk, how many recidivated? • Predictive value: TP / (TP + FP) • Defendant: what's the probability I'll be incorrectly classified as high risk? • False positive rate: FP / (FP + TN) • Society: is the selected set demographically balanced? • Demographic parity 39 Aravind Narayanan - 21 Fairness Definitions
  • 40.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Group Fairness • Do outcomes differ between groups (e.g. demographic), which we have no reason to believe are actually different? • Motivation of the Pro-Publica article 40 Aravind Narayanan - 21 Fairness Definitions Gbmtf! Qptjujwf Gbmtf! Ofhbujwf
  • 41.
    Full Stack DeepLearning - UC Berkeley Spring 2021 No "correct" group fairness definition 41 Aravind Narayanan - 21 Fairness Definitions
  • 42.
    Full Stack DeepLearning - UC Berkeley Spring 2021 It gets worse... • Okay, then let's just pick most important two metrics (FPR and FNR), and allow the model to use protected class attributes. • Now we fail individual fairness -> • Fine, let's just pick one! (e.g. equal FPR) • We still sacrifice some utility (e.g. public safety, or number of defendants released) 42 Aravind Narayanan - 21 Fairness Definitions
  • 43.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Interactive Example 43 https://research.google.com/bigpicture/attacking-discrimination-in-ml/
  • 44.
    Full Stack DeepLearning - UC Berkeley Spring 2021 By the way... 44 Aravind Narayanan - 21 Fairness Definitions
  • 45.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Trade-offs 45 Aravind Narayanan - 21 Fairness Definitions
  • 46.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Seeing the water 46
  • 47.
    Full Stack DeepLearning - UC Berkeley Spring 2021 47 https://www.mobilizegreen.org/blog/2018/9/30/environmental-equity-vs-environmental-justice-whats-the-difference
  • 48.
    Full Stack DeepLearning - UC Berkeley Spring 2021 48 https://www.mobilizegreen.org/blog/2018/9/30/environmental-equity-vs-environmental-justice-whats-the-difference
  • 49.
    Full Stack DeepLearning - UC Berkeley Spring 2021 49 https://www.mobilizegreen.org/blog/2018/9/30/environmental-equity-vs-environmental-justice-whats-the-difference
  • 50.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Questions? 50
  • 51.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Representation 51
  • 52.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Representation 52 https://twitter.com/nke_ise/status/897756900753891328
  • 53.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Not a new problem, sadly 53
  • 54.
    Full Stack DeepLearning - UC Berkeley Spring 2021 ...or a problem in just our field 54 Recent improvement!
  • 55.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Large part of the solution 55 https://www.nytimes.com/2021/03/15/technology/artificial-intelligence-google-bias.html
  • 56.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Another example 56 Recent improvement!
  • 57.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Translation 57 Aravind Narayanan - 21 Fairness Definitions Recent improvement!
  • 58.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Word Embeddings • Word2Vec introduced vector math on word embeddings • Reveal harmful biases encoded in our language corpora • Potential solution: de-bias at training time, but at least make user aware 58
  • 59.
    Full Stack DeepLearning - UC Berkeley Spring 2021 GPT-3 59
  • 60.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Recent Language Models • Paper apparently led to authors being fired by Google 60 http://faculty.washington.edu/ebender/papers/Stochastic_Parrots.pdf
  • 61.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Seeing the water • Whether language models should reflect the world as it is in the data, or as society believes it should be, depends on what use they are applied to. • And when we do want them to reflect the world as it should be, do we agree on what that is? 61
  • 62.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Face Recognition • Old context: no expectation of privacy in public • New context: "in public" now means "on any street in any city, or on any website on the internet" • Is it ethical to work on this? • Is it a problem if it does not work as well on some ethnicities? 62 https://fairmlbook.org/introduction.html
  • 63.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Questions? 63
  • 64.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Best practices 64
  • 65.
    Full Stack DeepLearning - UC Berkeley Spring 2021 What do practitioners need? 65 • support in fairness-aware data collection and curation • overcoming teams’ blind spots • implementing more proactive fairness auditing processes • auditing complex ML systems • deciding how to address particular instances of unfairness • addressing biases in the humans embedded throughout the ML development pipeline 2019
  • 66.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Some suggestions • Ethical risk sweeping
 treat like cybersecurity penetration testing • Expanding the ethical circle
 whose interests, desires, experiences, values have we just assumed instead of consulted? • Think about the terrible people
 Who might abuse, steal, weaponize what we build? What incentives are we creating? • Closing the loop
 Remember that this is not a process to complete and forget. Set up ways to keep improving. 66 https://www.youtube.com/watch?v=av7utkFXbU4
  • 67.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Model Cards 67 https://modelcards.withgoogle.com/face-detection
  • 68.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Bias Audit 68 http://www.datasciencepublicpolicy.org/projects/aequitas/
  • 69.
    Full Stack DeepLearning - UC Berkeley Spring 2021 69 http://www.datasciencepublicpolicy.org/projects/aequitas/
  • 70.
    Full Stack DeepLearning - UC Berkeley Spring 2021 70 AI bias is rooted in societal inequities, manifested in data, and amplified by AI. Unbiased societal AI is an impossible goal for a single project to achieve, but greatly mitigating its harmful effects is not. Visualization courtesy of Eric Wang
  • 71.
    Full Stack DeepLearning - UC Berkeley Spring 2021 71 Build a diverse AI team to bring lived experience to their work Display insights about bias contextually and intuitively in product Make AI fairness and transparency a key component of our work Visualization courtesy of Eric Wang
  • 72.
    Full Stack DeepLearning - UC Berkeley Spring 2021 A professional code of ethics? 72
  • 73.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Medicine 73
  • 74.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Armed Forces 74
  • 75.
    Full Stack DeepLearning - UC Berkeley Spring 2021 75
  • 76.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Questions? 76
  • 77.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Learn more 77
  • 78.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Experts • Practical Data Ethics course from Rachel Thomas: https://ethics.fast.ai • Single lecture from Fast.ai 2020 course: https:// www.youtube.com/watch?v=krIVOb23EH8 78
  • 79.
    Full Stack DeepLearning - UC Berkeley Spring 2021 79 https://fairmlbook.org
  • 80.
    Full Stack DeepLearning - UC Berkeley Spring 2021 80 https://dssg.github.io/fairness_tutorial/
  • 81.
    Full Stack DeepLearning - UC Berkeley Spring 2021 81
  • 82.
    Full Stack DeepLearning - UC Berkeley Spring 2021 82
  • 83.
    Full Stack DeepLearning - UC Berkeley Spring 2021 Thank you! 83