New article out - written with @p_alapietila - in which we put forward a comprehensive framework for Global Cooperation on #AI & its #Governance, focusing on the Why, What and How of global cooperation.

Preprint here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3696519.

Main points in thread(1/12)

There is a growing awareness of the need to cooperate & many initiatives took off. Why? Esp. important to shift global mindset from zero sum game / protectionist approach (with winners v. losers but in long term, everyone loses) towards infinite game (all can/should gain). (2/12)

What challenges to address? Protect citizens from AI risks across the globe & secure conditions for healthy & fair global competition, so as to enable socially beneficial AI innovation. This requires the creation of a holistic level playing field, along two dimensions: (3/12)

(a) Horizontal: common requirements must render entire socio-technical environment of AI systems trustworthy. Beyond AI systems, must also include environment of data & infrastructure. No TW system w/o TW data & TW infrastructure (= the system-data-infrastructure trinity). (4/12)

(b) Vertical: cooperation also needed on domain-specific issues in tailored way: areas/applications where AI’s benefits for humanity can be maximised (AI4Good), where there are significant risks (military, misinformation, mass surveillance) & ad hoc matters (e.g. Covid). (5/12)

As to the How of global cooperation, we have 7 points of advice (yes, we like that number :-)).
1. Holistic approach is crucial, but so is the need for speed & context-specificity. These 3 aspects are not always in harmony and must be balanced depending on issue at stake; (6/12)

2. The rules of engagement amongst cooperation partners must be clarified in advance, especially when they compete with each other on values rather than (only) on markets; (7/12)

3. Cooperation via existing fora is to be favored over creating new ones, unless justification;
4. When cooperation mandates of different fora overlap they should create a network of networks to exchange knowledge & build collective wisdom (no competition -> coopetition); (8/12)

5. Considering models of differentiated cooperation is useful, so that cooperation partners can cooperate at different speeds based on value-alignment / shared objectives. Yet the ultimate aim should be an as broad as possible coalition; (9/12)

6. An inclusive and transparent way of working is needed amongst cooperation partners, mindful of power imbalances. Not everyone has easy access to a seat at the cooperation table, but all voices should be heard; (10/12)

7. Feedback loops are needed to improve the cooperation approach along the way, informed by science/evidence. Simultaneously, we need to prepare & empower the next generation of global cooperators through broader education on the capabilities & limitations of AI. (11/12)

Establishing meaningful global cooperation on AI & its governance is a #marathon, not a sprint. Lots of work to do, so global cooperation partners better roll up their sleeves!

For more details, read the article here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3696519. (12/12)