The End Of Knowledge - Vault 4/27 - AI For Good - 2019 - International Committee of the Red Cross

graph LR classDef icrc fill:#d4f9d4, font-weight:bold, font-size:14px classDef dataPrivacy fill:#f9d4d4, font-weight:bold, font-size:14px classDef bigData fill:#d4d4f9, font-weight:bold, font-size:14px classDef conflictTech fill:#f9f9d4, font-weight:bold, font-size:14px classDef governance fill:#f9d4f9, font-weight:bold, font-size:14px A[International Committee of
the Red Cross] --> B[Works for International
Committee of Red Cross. 1] A --> C[ICRC develops humanitarian law
AI changed war. 2] A --> D[AI potential for SDGs,
consider risks. 3] A --> E[Visits prisoners, conducts
private interviews. 4] A --> F[ICRC promises data privacy,
agreement required. 5] F --> G[Data privacy harder today
despite history. 6] A --> H[Rethink data protection
in humanitarian environments. 7] A --> I[People worldwide more
aware of data. 8] A --> J[Data management critical
in violent environments. 9] A --> K[ICRC challenged
on data management. 10] K --> L[Trust crucial
for data collection. 11] A --> M[Uses big data, algorithms
to assess needs. 12] M --> N[Analyzes tweets for
humanitarian response. 13] M --> O[Reflect on algorithm biases,
impartiality. 14] A --> P[Integrate humanitarian principles
in technology design. 15] A --> Q[AI must consider
modern conflict weapons. 16] Q --> R[Modern wars use tech,
autonomous weapons. 17] Q --> S[Autonomous weapons may target
without human control. 18] S --> T[Such weapons make decisions
on combatants. 19] S --> U[Autonomous mass destruction weapons
accessible to non-state actors. 20] A --> V[Key: how much
human control remains. 21] V --> W[Human control, agency
in weapons essential. 22] A --> X[Only humans accountable
for war decisions. 23] A --> Y[Maintaining human accountability
in war critical. 24] A --> Z[Reflect on needed dialogue
for AI governance. 25] Z --> AA[States struggle to regulate AI,
autonomous weapons. 26] AA --> AB[Regulating AI, developing
common rules challenging. 27] A --> AC[Need ethical AI regulation
in critical areas. 28] A --> AD[Urgent consideration of AIs
impact on people. 29] A --> AE[AI for good immense,
mitigate risks. 30] class A,B,C,D,E,H,P,V,X,Y governance class F,G dataPrivacy class K,L bigData class Q,R,S,T,U conflictTech class Z,AA governance

Resume:

1.- The speaker works for the International Committee of the Red Cross (ICRC), focusing on people affected by war, violence, and disasters worldwide.

2.- The ICRC has a mandate to develop international humanitarian law, which regulates war. AI and data have already changed this landscape.

3.- There is great potential for AI to contribute to the Sustainable Development Goals, but it's important to consider the risks and questions.

4.- The ICRC visits 700,000-800,000 prisoners worldwide as part of its mandate. When visiting, they conduct private interviews to collect personal data.

5.- There is a social contract that the ICRC will treat the prisoners' data with integrity and not share it without agreement.

6.- Guaranteeing data privacy and control is more difficult today than 100 years ago, despite the ICRC's long history of data collection.

7.- Organizations must rethink how to protect people's data and be aware of data rights and standards, especially in humanitarian environments.

8.- In recent years, people worldwide, even in complex environments, are increasingly connected via mobile phones and aware of data importance.

9.- In violent environments, the ability to manage one's own data can be a matter of life and death. People scrutinize data management.

10.- The ICRC is increasingly challenged on data management. People want an organization as good as Google at managing private crisis data.

11.- The ICRC's reputation for generating trust will be critical in attracting people to provide data, especially in insecure situations.

12.- The ICRC has begun using big data and algorithms to improve its ability to quickly assess needs on the ground.

13.- Algorithmic analysis of data like tweets has helped identify issues like water, sanitation, and health problems to improve humanitarian response.

14.- However, the ICRC had to reflect on what algorithms to use and potential biases, working with academics to construct impartial algorithms.

15.- Integrating humanitarian principles into technology design was necessary even for relatively simple applications of data analysis for needs assessments.

16.- In considering AI, humanitarian organizations must also think about the weapons used in modern conflicts, which increasingly involve technology.

17.- Modern wars are often fought with special forces, proxies, mercenaries, cyber capabilities, drones, and increasingly, autonomous weapons.

18.- In the next 5-7 years, very small autonomous weapons may be able to enter a room and select targets without human control.

19.- These weapons could make decisions about combatants vs. non-combatants based on parameters like age and appearance. The example given was targeting old, bald men.

20.- Autonomous weapons with mass destruction impact will likely become relatively inexpensive and accessible to non-state actors, not just governments.

21.- The key question is how much human control and agency will remain at the core of these AI-enabled weapons systems.

22.- The ICRC believes that regardless of technological developments, human control and agency in weapons systems and targeting decisions is essential.

23.- Weapons, as machines, cannot be legally accountable - only human beings can and must be held accountable for decisions in war.

24.- Maintaining human accountability in war is already difficult but absolutely critical. Outsourcing kill decisions to AI is hugely concerning.

25.- The speaker wants to reflect with the audience on what dialogue is needed regarding AI development and governance.

26.- States are struggling to build consensus and regulate AI development, including in the context of autonomous weapons.

27.- Key questions include how to regulate AI, who is in charge of developing common understandings and rules.

28.- Some say significant regulation and examination of ethical questions is needed for AI in areas like healthcare that impact human beings.

29.- The speaker argues that for the sake of humanity and the Sustainable Development Goals, we must urgently consider AI's impact on people, starting with weapons.

30.- In summary, the potential of AI for good is immense, but mitigating risks, maintaining human agency, and building governance is essential.

Knowledge Vault built byDavid Vivancos 2024