AI R&D GUIDELINES. Oct , 2017, Paris. AI R&D Guidelines

Transcription

1 AI R&D GUIDELINES Oct , 2017, Paris

2 G7 ICT MINISTERS MEETING (Takamatsu, Apr. 2016) - Former Minister Sanae TAKAICHI of Japan proposed that G7 countries take the lead in international discussions taking into consideration the formulation of *, which consists of AI R&D Principles, as a non-regulatory and non-binding international framework. - The participating countries agreed to her proposal. *A preliminary proposal on including 8 principles was distributed before TAKAICHI s proposal. 2

3 CONFERENCE TOWARD AI NETWORK SOCIETY, MIC (Tokyo, Oct to Present) Has discussed and prepared the draft of the Guidelines for developers of AI Systems. MIC HQs, Kasumigaseki, Tokyo Published in July, this year. /main_content/ pdf. Began to discuss what are expected to be taken into consideration in use of AI Systems, last Thursday. (Picture in RH bottom) 3

4 Draft AI R&D GUIDELINES Tentative Translation Draft * - Draft guidelines for developers of AI Systems - Prepared for international discussions (e.g., G7, OECD) - Compiling AI R&D [9] Principles (with comments) Purpose * To increase benefits and mitigate risks of AI systems thru the sound progress of AI networks. Protecting users interests; deterring spread of risks Realizing a human-centered society ( Wisdom Network Society ) 4

5 Draft AI R&D GUIDELINES (cont d) Tentative Translation Basic Philosophies To achieve a human-centered society. To share the Guidelines as non-binding soft law and their best practices among stakeholders internationally. To ensure a balance between benefits & risks of AI networks. To ensure technological neutrality; and to avoid excessive burden on developers. To review constantly and revise flexibly as necessary the Guidelines. 5

6 AI R&D [9] PRINCIPLES Tentative Translation To promote the benefits of AI systems To mitigate the risks of AI systems II. Transparency IV. Safety VI. Privacy I. Collaboration III. Controllability V. Security VII. Ethics To improve acceptance by users and other stakeholders VIII. User Assistance IX. Accountability 6

7 Principle of: Developers should: I. Collaboration Pay attention to the interconnectivity and interoperability of AI systems. II. Transparency AI R&D [9] PRINCIPLES Tentative Translation Pay attention to the verifiability of inputs/outputs of AI systems and explainability of their decisions. III. Controllability IV. Safety Pay attention to the controllability of AI systems. Take it into consideration that AI systems will not harm the life, body, or property of users or third parties through actuators or other devices. V. Security Pay attention to the security of AI systems. VI. Privacy VII. Ethics VIII. User Assistance IX. Accountability Take it into consideration that AI systems will not infringe the privacy of users or third parties. Respect human dignity and individual autonomy in R&D of AI systems. Take it into consideration that AI systems will support users and make it possible to give them opportunities for choice in appropriate manners. Make efforts to fulfill their accountability to stakeholders including users of AI systems. 7

8 THE MOTORCYCLE PROBLEM Locked-on target!!! AV (2) (1) w/o a helmet wearing a helmet Choice (1) could minimize loss. But it would punish the recommendable rider (1). We might have to consider at least the following Principles of: IV. Safety; VIII. Ethics; IX. User Assistance; and X. Accountability. Drawn by Hirano based upon hypos. in Noah J. Goodall, Can You Program Ethics into a Self-Driving Car?, IEEE SPECTRUM, May 31, 2016, available at < (last visited Oct. 22, 2017); Patrick Lin, The Robot Car of Tomorrow May Just Be Programmed to Hit You, WIRED, May 6, 2014, available at < (last visited Oct. 22, 2017). 8

9 Thank you for your attention! ;-) 9