Knowledge Vault 4 /57 - AI For Good 2020
What is AI for Good anyway?
Sasha Luccioni
< Resume Image >
Link to IA4Good VideoView Youtube Video

Concept Graph & Resume using Claude 3 Opus | Chat GPT4o | Llama 3:

graph LR classDef intro fill:#f9d4d4, font-weight:bold, font-size:14px classDef sasha fill:#d4f9d4, font-weight:bold, font-size:14px classDef caution fill:#d4d4f9, font-weight:bold, font-size:14px classDef recommend fill:#f9f9d4, font-weight:bold, font-size:14px classDef gender fill:#f9d4f9, font-weight:bold, font-size:14px classDef food fill:#d4d4f4, font-weight:bold, font-size:14px classDef pandemic fill:#d4f4f9, font-weight:bold, font-size:14px A[What is AI
for Good anyway?] --> B[Xenia introduces
AI for
Good Breakthrough. 1] A --> C[Andrew moderates,
introduces keynote
Sasha. 2] C --> D[Sasha questions
what makes
AI good. 3] C --> E[AI can address
challenges,
intentions insufficient. 4] E --> F[Success: Stanford,
Sustainability Network,
COVID-19 tool. 5] E --> G[Caution: COMPAS,
UK exams,
anti-poaching. 6] E --> H[Pitfalls: magical thinking,
experimenting on
vulnerable. 7] E --> I[Recommendations: right questions,
right people,
transparency. 8] I --> J[Ask about real problem,
data source. 9] I --> K[Include, empower
stakeholders, engage
diverse voices. 10] I --> L[Be transparent about
risks, benefits,
bias. 11] I --> M[AI for good needs
cross-discipline
collaboration. 12] B --> N[Hanin introduces
Gender Equity
track teams. 13] N --> O[Roberta presents UNs
judicial gender
bias project. 14] O --> P[Judicial AI gender
bias risks
inequality. 15] P --> Q[Solution: global guidelines,
practice repository,
guide. 16] Q --> R[Collaborators needed for
research, testing,
dissemination. 17] B --> S[Caroline introduces Future
of Food
track teams. 18] S --> T[Maurice presents Grow Nexts
growth modules. 19] T --> U[Grow Next optimizes
food growth
recipes. 20] U --> V[Team seeks expertise,
partners, aims
accessibility. 21] B --> W[Andrew introduces Pandemic
track team
COVID Sense. 22] W --> X[Leontius explains COVID
Senses data
collection. 23] X --> Y[Federated learning for
privacy, needs
ambassadors. 24] B --> Z[Hanin re-introduces
Hannah on NLP
gender bias. 25] Z --> AA[NLP gender biases
harm, balanced
data helps. 26] AA --> AB[Team detects gender biases,
expands scope. 27] B --> AC[Caroline re-introduces Hameds
satellite data
approach. 28] AC --> AD[Hamed explains efficient
crop growth
necessity. 29] AD --> AE[Radiant Earth uses AI
for sustainable
farming. 30] class A,B intro class C,D,E,F,G sasha class H,I,J,K,L,M recommend class N,O,P,Q,R gender class S,T,U,V food class W,X,Y pandemic class Z,AA,AB gender class AC,AD,AE food

Resume:

1.- Xenia Fonten from ITU introduces the AI for Good Breakthrough Days, which aims to identify practical AI applications to advance the SDGs.

2.- Andrew Tate from X Prize moderates and introduces keynote speaker Sasha Luccioni from Mila AI Institute.

3.- Sasha explores what makes AI "good" - is it the application domain, problem solved, positive impact, intentions, or all of these?

4.- AI holds potential to address major challenges like poverty, food scarcity and climate change, but good intentions alone aren't enough.

5.- Success stories: Stanford's bottom-up approach, Computational Sustainability Network's top-down method, and van der Scharz group's rapid COVID-19 tool deployment.

6.- Cautionary tales: COMPAS judicial system bias, UK exam result prediction bias, Africa anti-poaching facial recognition flaws.

7.- Pitfalls: viewing AI as magical fairy dust for positive impact without concrete solutions, using AI to experiment on vulnerable populations.

8.- Recommendations: Ask the right questions, include the right people, be transparent about risks/benefits/limitations.

9.- Ask if we're solving the real problem or a proxy, if AI is the right solution, where data comes from.

10.- Include stakeholders, empower people who will use the AI, engage diverse voices/ideas beyond typical Western AI hubs.

11.- Be transparent about risks/benefits, scope/limitations, how bias is checked, what happens with false positives, who makes final decisions.

12.- AI for good requires more than technology - it takes collaboration across disciplines, social engagement, policy changes.

13.- Hanin Kaluff introduces the two Gender Equity track teams: Global Judicial Integrity Network and Umeå/Uppsala Universities.

14.- Roberta Solis presents the UN project to develop global recommendations for judiciaries to address gender bias in AI systems.

15.- Gender bias in judicial AI risks unequal access to justice and fair trials for women, especially in gender-based violence cases.

16.- Solution: global guidelines, repository of practices, self-assessment guide for judiciaries. Dissemination through UN channels and partners is key.

17.- Collaborators needed for research, AI/gender expertise, pilot testing, dissemination. Guidelines aim to become international standards adopted by judiciaries.

18.- Caroline Kolta introduces the two Future of Food track teams: Grow Next and Radiant Earth Foundation.

19.- Maurice Zundars presents Grow Next's semi-autonomous growth modules using IoT, AI and apps to efficiently grow fresh food in megacities.

20.- Grow Next uses digital crop growth recipes to optimize nutrition, safety, freshness and taste. B2B then B2C go-to-market strategy.

21.- Team seeks expertise in engineering, plant science, analytics, and local partners in megacities. Goal to increase access to healthy food.

22.- Andrew Tate introduces the Pandemic track team COVID Sense, using smartphone data and AI to detect early COVID-19 symptoms.

23.- Leontius Tyriades explains how COVID Sense collects metadata, self-reported symptoms, breathing/cough sounds, heart rate for AI risk assessment.

24.- Federated learning protects data privacy. Aims to empower people for health self-management and guide policymakers. Seeking country ambassadors and collaborators.

25.- Hanin re-introduces Hannah Davini presenting the Umeå/Uppsala team's work on detecting gender bias in natural language processing training data.

26.- NLP systems exhibit gender biases (e.g. in translation, job applicant filtering) that harm people. More balanced data can help.

27.- Team developed methods to surface implicit gender biases in training data for English/Swedish. Expanding to more languages/social dimensions needed.

28.- Caroline re-introduces Hamed Mohammed presenting Radiant Earth's satellite data and AI approach to map crops and optimize food production.

29.- Hamed explains the need to grow crops more efficiently to meet demands while reducing environmental damage. Global solutions are needed.

30.- Radiant Earth uses satellite imagery, cloud computing, AI/ML to map croplands and generate future optimal sustainable farming scenarios. Seeking funding and partners.

Knowledge Vault built byDavid Vivancos 2024