The End Of Knowledge - Vault 4/20 - AI For Good - 2018

graph LR classDef summit fill:#d4f9d4, font-weight:bold, font-size:14px classDef concerns fill:#f9d4f9, font-weight:bold, font-size:14px classDef applications fill:#f9e1d9, font-weight:bold, font-size:14px classDef impact fill:#d9e6f2, font-weight:bold, font-size:14px A[Transformations on the
Horizon] --> B[Last summit was
inspiring but naive. 1] B --> C[AI weaponization, bias,
transparency concerns arise. 2] C --> D[Concerns about data ownership,
responsibilities emerge. 3] A --> E[Second summit focuses on
protection, consequences. 4] A --> F[Balance AI benefits, mitigate
societal risks. 5] A --> G[Outwardly vs. inwardly
turning AI for good. 6] G --> H[Focus on helping
vulnerable populations. 7] H --> I[Simple applications may
be most powerful. 8] H --> J[Example: Pula insures
African farmers. 9] J --> K[Policies via cell phones,
satellite data. 10] K --> L[Small application vital
for crop failures. 11] A --> M[Tech helps meet all
sustainability goals. 12] M --> N[AI amplifies other tech,
brings efficiencies. 13] M --> O[Tech progress may worsen
inequality, unemployment. 14] O --> P[Automation pressures job
creation, wage growth. 15] P --> Q[Major disruptions before AI
creates jobs. 16] Q --> R[Productivity gains favor
the well-off. 17] R --> S[Vulnerable affected by
lack of work, inequality. 18] A --> T[AI may dehumanize,
hinder human flourishing. 19] T --> U[Mitigate harms,
address societal impacts. 20] U --> V[Harms from bad design,
misappropriation, failures. 21] V --> W[Ethical, governance concerns
need addressing. 22] W --> X[Solutions: tech fixes,
better corporate oversight. 23] A --> Y[Balance outwardly and inwardly
turning AI. 24] Y --> Z[Suggestions for better
phrasing invited. 25] class B,C,D,E,F summit class G,H,I,J,K,L,M,N,O,P,Q,R,S concerns class T,U,V,W,X applications class Y,Z impact

Resume:

1.- Last year's AI for Good Global Summit was inspiring but seemed naive at times, especially regarding data use and human rights.

2.- Since then, issues like AI weaponization, behavior manipulation, algorithmic bias, and lack of transparency have come to the fore.

3.- There are concerns around data ownership, rights, and responsibilities of those who control data.

4.- The second gathering shows more focus on protecting each other and potential negative consequences.

5.- The main question is how to reap AI benefits while mitigating risks and undesirable societal consequences.

6.- The talk distinguishes between outwardly turning AI for good (discrete benefits from specific applications) and inwardly turning AI for good.

7.- Outwardly turning AI for good should be the primary focus, particularly for helping vulnerable populations.

8.- Some of the most powerful applications may be simple ones, not just complicated future speculations.

9.- Example: Pula, an insurance company providing policies to tiny farmers in Africa, packaged with seed and fertilizer purchases.

10.- Policies are registered via ubiquitous cell phones, and claims are automatically processed using satellite data and deep learning algorithms.

11.- This small application is tremendously important, as crop failures occur frequently in these regions.

12.- Emerging technologies can help meet all 17 sustainable development goals, though not as the sole solution.

13.- AI may not be central to each goal but can amplify other technologies and bring efficiencies.

14.- However, progress in areas like decent work (Goal 8) and reduced inequality (Goal 10) may be exacerbated by tech progress.

15.- Technological unemployment and automation are already putting downward pressure on job creation and wage growth.

16.- Major disruptions are likely before AI potentially creates more jobs than it eliminates in the long run.

17.- The tech economy is designed so that productivity gains flow more towards the already well-off.

18.- Lack of work and inequality most affect the most vulnerable populations, impacting goals of no poverty (Goal 1) and zero hunger (Goal 2).

19.- Progress in AI may not always serve sustainability goals and can dehumanize people, rob them of meaning, and interfere with human flourishing.

20.- Inwardly turning AI for good involves mitigating harms, addressing potential negative consequences, and examining societal impacts.

21.- Harms can come from system failures due to bad design, incompetence, underestimating low-probability high-impact events, and misappropriation by bad actors or elites.

22.- We need to look at ethical and governance concerns, bringing appropriate oversight and addressing gaps.

23.- Solutions can involve technological fixes, better corporate oversight, or other mechanisms.

24.- The talk emphasizes the importance of considering both outwardly and inwardly turning AI for good.

25.- The speaker invites suggestions for better phrasing to make this distinction.

Knowledge Vault built byDavid Vivancos 2024