Deeplearning - Ai Deeplearning - Ai
Deeplearning - Ai Deeplearning - Ai
Deeplearning - Ai Deeplearning - Ai
DeepLearning.AI makes these slides available for educational purposes. You may not use or
distribute these slides for commercial purposes. You may make copies of these slides and
use or distribute them for educational purposes as long as you cite DeepLearning.AI as the
source of the slides.
Andrew Ng
AI and society
• AI and hype
• Limitations of AI
-Bias AI and Ethics
-Adversarial attacks
Andrew Ng
AI and Society
deeplearning.ai A realistic view of AI
Andrew Ng
Goldilocks rule for AI
Andrew Ng
Limitations of AI
• Performance limitations
• Explainability is hard (but sometimes doable)
Right-sided
Pneumothorax
(collapsed lung)
[Rajpurkar et al. (2018). CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays with Deep Learning.]
[Wang et al. (2017). ChestX-ray8: Hospital-scale Chest X-ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases. IEEE CVPR]
[Images source: NIH Clinical Center Image dataset: https://nihcc.app.box.com/v/ChestXray-NIHCC]
Andrew Ng
Limitations of AI
Andrew Ng
AI and Society
deeplearning.ai Discrimination / Bias
Andrew Ng
AI learning unhealthy stereotypes
• Man : Woman as Father : Mother
• Man : Woman as King : Queen
• Man : Computer programmer as Woman : Homemaker
Computer programmer
Homemaker
Man: (1,1) 4
Woman
Computer programmer: (3,2) 3
2 Computer programmer
Woman: (2,3)
1
Homemaker: (4,4) Man
1 2 3 4
Bolukbasi et al. (2016). Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings.
Andrew Ng
Why bias matters
• Hiring tool that discriminated against women
• Facial recognition matching dark skinned
individuals to criminal mugshots
• Bank loan approvals
• Toxic effect of reinforcing unhealthy stereotypes
Andrew Ng
Combating bias
• Technical solutions:
- E.g., “zero out” the bias in words
- Use less biased and/or more inclusive data
• Transparency and/or auditing processes
• Diverse workforce
- Creates less biased applications
Andrew Ng
AI and Society
deeplearning.ai Adversarial attacks
on AI
Andrew Ng
Adversarial attacks on AI
Minor perturbation
Hummingbird Hammer
Minor perturbation
Hare Desk
Andrew Ng
Physical attacks
Andrew Ng
Physical attacks
[Sharif et al. (2016). Accessorize to a Crime: Real and Stealthy Attacks on State-of-the-Art Face Recognition.]
[Eykholt et al. (2018). Physical Adversarial Examples for Object Detectors.]
[Brown et al. (2018). Adversarial Patch.]
Andrew Ng
Adversarial defenses
• Defenses do exist, but incur some cost
• Similar to spam vs. anti-spam, we may be in an
arms race for some applications
Andrew Ng
AI and Society
deeplearning.ai Adverse uses of AI
Andrew Ng
Adverse uses of AI
• DeepFakes
- Synthesize video of people doing things they never did
Andrew Ng
AI and Society
deeplearning.ai AI and developing economies
Andrew Ng
Developing economies
“leapfrog”
-Mobile phones
-Mobile payments
-Online education
Andrew Ng
How developing economies can build AI
• US and China are leading, but all AI
communities are still immature
• Focus on AI to strengthen a country’s vertical
industries
• Public-private partnerships to accelerate
development
Andrew Ng
AI and Society
deeplearning.ai AI and jobs
Andrew Ng
AI’s impact on jobs worldwide
Andrew Ng
AI’s impact on jobs worldwide
Andrew Ng
Some solutions
Andrew Ng
AI for Everyone
deeplearning.ai
Conclusion
Andrew Ng
What you’ve learned
• What is AI?
• Building AI projects
• Building AI in your company
• AI and society
Keep learning!
• Online courses, books, blogs, ...
• deeplearning.ai mailing list
Andrew Ng
Andrew Ng