Ethics of AI - IIT Delhi
Transcript of Ethics of AI - IIT Delhi
![Page 1: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/1.jpg)
Ethics of AI
Mausam
![Page 2: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/2.jpg)
Topics
• Robots vs Humans
• Jobs
• Bias
• Fairness
• Accountability
• Transparency
• Privacy
• Ethical uses
![Page 3: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/3.jpg)
![Page 4: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/4.jpg)
![Page 5: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/5.jpg)
Future of Jobs
• AI Present– 40% of companies struggle to hire and retain data scientists
– ~1/3rd of the top 400 companies lack SoA data analysis tools and personnel
– 364K new jobs expected by 2020. • (50K currently vacant in India)
• ~1/3rd of jobs could be replaced by 2030– (many different reports)
• AI will create more jobs than it eliminates– (Gartner report)
• Teams of AI + Human Intelligence will be common
![Page 6: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/6.jpg)
Key Challenge: Robustness
![Page 7: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/7.jpg)
Key Challenge: Data Bias
he:she
surgeon:nursebrilliant:lovelyarchitect:interior designer
![Page 8: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/8.jpg)
Key Challenge: Fairness
![Page 9: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/9.jpg)
Key Challenge: Transparency
• Almost no idea why Deep Learning
– works, or
– doesn’t work
• Important for human-AI teams
• New research agenda: Xplainable AI
![Page 10: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/10.jpg)
Key Challenge: Accountability
• Who/What is responsible?
– Company who designed the car
– Engineer who designed the ML algorithms
– Owner who bought the car
– Driver who drove the car and gave training data
![Page 11: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/11.jpg)
Key Challenge: Privacy
![Page 12: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/12.jpg)
Key Challenge: Human-AI Interaction
• Defining the objective function
– “You should not see any dirt”
– “Have no dirt”
– “If there is dirt, clean the dirt”
• Cognitive Science + AI
– Understanding humans and communicating w them
![Page 13: Ethics of AI - IIT Delhi](https://reader030.fdocuments.us/reader030/viewer/2022012716/61ae876a30dc6a373e5a16d0/html5/thumbnails/13.jpg)
Ethical uses of AI
• Dynamite vs. bomb
• Intelligent weapons?
– reduce barrier to wars
– kill targeted people
– democratize weapons
• Automated doctor?