Regulating Artificial Intelligence
1. Regulating Artificial Intelligence Copyright © 2015 Orren Prunckun. All Rights Reserved.
2. What Is Artificial Intelligence Technology?
3. “A software that can perform tasks that require human intelligence or cognitive functions.” “An Intelligent machine or agent.”
4. • Reasoning and problem solving. • Knowledge representation. • Planning and problem solving. • Learning (AKA Machine Learning). • Natural language processing. • Visual and audible Perception (Machine perception, Computer vision, and Speech recognition). • Motion and manipulation (Robotics). • Social intelligence. • Abstract and creative thinking (Computational creativity). • General intelligence and decision making (General AI). • Many, many more… *Italics for common usage of terms. Cognitive Functions…
5. The Turing Test • Widely used as the test for Artificial Intelligence. • Developed by Alan Turing in 1950. • Predicts a machines ability to exhibit intelligent behaviour. • In other words, convince a human that the machine itself if is a human also. • Yet the test is only done through responses to written questions. • General Artificial Intelligence (AKA Strong AI) can pass Turing Test. • Weak Artificial Intelligence (AKA Narrow AI) cannot pass the Turing Test.
6. Uses of AI… Categories: • Search and optimization. • Logic. • Reasoning. • Learning (Deep learning.) • Neural networks( models human brain). • Control. • Language. • Evaluation. • Many, many more… *Italics for common usage of terms.
7. What Is The History Of Artificial Intelligence Technology?
8. History Of Artificial Intelligence Technology… • Stories and fiction. • Philosophers and mathematicians. • Discoveries in neuroscience, psychology, linguistics and cybernetic = potential for an electronic brain. • Advances in computer science • First AI designed in 1943. • Replication of knowledge and analytical in 1980s. • First self driving cars in 1980s. • Deep Blue won chess in 1997. • IBM Watson in Jeopardy in 2011.
9. What Are Some Types Of Artificial Intelligence Technology?
10. Types Of Artificial Intelligence… Practical examples: • Health care. • Finance. • Marketing. • Video games. • Automotive (Self Driving Cars – today!) • Many, many more…
11. Why Is Artificial Intelligence Technology Important For Society?
12. Importance Of Artificial Intelligence… Not mainstream yet, but emerging. • Possibility of taking over existing jobs ? • Possibility of creating new jobs and job types? • Compliment our jobs not supplement them? • Ability to change society and thus laws? • The Singularity?… • …and human apocalypse (OpenAI a non-profit AI research company, discovering and enacting the path to safe artificial general intelligence)? • Enhance not replace humans? • What else?
13. Could A Robot Do Your Job? Probably!
14. Based on this as a soon-to-be-grad, what will you do to future-proof yourself against AI?
15. Case Study – Self Driving Cars
16. Who Are The Various Stakeholders That Influence The Development Of Legal Regulation Of Self Driving Cars?
17. Stakeholders That Influence The Development Of Regulation For Self Driving Cars… International Regulators • United Nations. Domestic Regulators • The Australian Executive (Department of Industry, Innovation and Science). • Parliament of Australia. Entrepreneurs, designers, manufacturers, developers & their legal entities • Google. • Facebook. • Apple. • Etc. Citizens of the World • International community. • Domestic community.
18. Stakeholders That Influence The Development Of Regulation For Self Driving Cars… Road users • Passengers. • By standers. Vehicle Owners • Taxi licences plate holders. • Public transport. Hackers Insurers
19. What Are The Current International And Domestic Frameworks That Regulate Self Driving Cars?
20. Recall that a Treaty is: “An agreement between States and Nations.”
21. • Not between Citizens or between a Nations States and Territories. • They serve as a way to practice stable and organized international relations. • Binding at international law. • A contract (contracts needs intent). • It is consent-based governance. • A State can only abide and enforce a treaty by if they consented. • If they didn’t consent then they can ignore it. • Ratification means confirmation. • This maintains State sovereignty (independence). • No State prosecutes treaties unless it’s after a major war like WWI or WWII. • Reality is national shaming, sections, diplomats etc. • Creates issues for Regulating Technology! Two types: • Bilateral treaty – between two States. • Multilateral treaty – between many States (UN Chatter).
22. Treaties help: • Create alliances in an interdependent, globalized, tech enabled work. • With international issues that cannot be solved by States alone. • Maintain State sovereignty (independence). • With Australian national interests. • Middle ranking power with finite negotiating resources. • Military and economical stake so we are not vulnerable. • Geo-isolation and population size means we benefit.
23. • Space. • Post. • Shipping. • Defence. • Nuclear non-proliferation. • Environment. • Civil aviation. • War. • Sea and maritime boundaries. • Human rights. • World heritage. • Terrorism. • Drug trafficking. • Border protection. • Refugees and asylum seekers. • International organisations. • Etc… Some areas regulated by Treaties:
24. Treaties are established (very simply) by: 1. Power to enter into treaties is granted under s51 and s61 of The Australian Constitution. 2. In the jurisdiction of Executive not Legislature (Parliament). • Signed then tabled at Parliament to discuss the benefits and effects of and obligations on Australia and required implementation. • Consultation with States and Territories, industry and other interest groups. 3. Ratification makes the treaty binding, but in domestically. 4. News domestic laws are not required. • If current legislation is adequate, then no domestic laws are created. • If current legislation is inadequate, then commonwealth to state laws are created. 5. Implementation through Executive action.
25. Universal Declaration of Human Rights… • Humans have non-binding rights. • Not a treaty, so it does not directly create legal obligations for countries. • Only an expression of the fundamental values which are shared by all members of the international community. • Similar rights granted by Magna Carta and the Rule of Law.
26. Current Domestic Regulatory Frameworks… Domestic (not to Australia) • Germany – a driver must have a “minimum” level attention. • California – a driver must monitor the safe operation. • Nevada – a driver must be behind the wheel.
27. Current Domestic Regulatory Frameworks… • What about the doctrine of Nulla poena sine lege – no penalty without a law? • Are self driving cars legal based on the principle of that everything is permitted unless prevented by law? • When laws cannot guide us, do we revert to ethics? • Is the law playing catch-up?
28. Ethics… • “Standards of behaviour that help us decision course of action for a situation.” • Not law, religion or morality. • Law and ethics don’t always line up. Consequentialist theories • The Utilitarian Approach • Benefits the largest group. • The Egoistic Approach • Appeals to self interest. • The Common Good Approach • Benefits people as a whole Non Consequentialist theories • The Duty-Based Approach • Proper intention from duty or universal law. • The Rights Approach • Protect the rights of those affected. • The Fairness or Justice Approach • All should be treated alike. • The Divine Command Approach • What god commands. Agent centred theories • The Virtue Approach • Considers human virtues (dignity, mercy, justice, humour etc.) • The Feminist Approach • Considers minorities.
29. Applying Ethics… One way: 1. Determine the issue or problem. 2. Determine the stakeholders. 3. Gather available information and facts. 4. Formulate possible options. 5. Make a decision and consider it. 6. Apply it (or present it.)
30. Source: https://en.wikipedia.org/wiki/File:Trolley_problem.png
31. Activity • You have been tasked with create new regulations for self driving cars and programming for accidents (The Trolley Problem). – A no win scenario, 1 vs many will die – Unfortunately it’s a matter of kill or let die. You don’t initiate it, or have any responsibility for it. • Represent each stakeholder group from before – Manufacturer, AI (Driver), Driver, Passenger, Bystander etc. 1. Determine your stakeholder groups interests (15 minutes). 2. Create a law based on International treaties and Australian domestic law that regulates self-driving cars taking into consideration your stakeholder groups interests (15 minutes). 3. Discuss if this is mutually beneficial for all stakeholder groups (15 minutes).
32. Source: https://en.wikipedia.org/wiki/File:The_trolley_problem.svg
33. More Trolley Problem Questions… • What if it is a adult vs child? • What if it is many adults vs child? • What if it is someone that “deserves” to die via capital punishment? • What if it is a passenger? • What if it is a passenger vs many people? • Does you answer change if it was you on the receiving end? • How do you achieve justice? • How about if it mirrored the instinctive but split second decision of a human and took the same action? • What about road conditions, obstacles, flow on effects?
34. More Questions… Liability • Liability means who’s is to blame? • How do you determine guilt? • What defines personhood and consciousness? – Natural birth, pain, memories, intention etc.? • Does Artificial Intelligence have personhood? • Will Artificial Intelligence have personhood?
35. Even More Questions… Fault and intention • Actus Reus mean action in criminal law. • Mens Rea mean intention in criminal law, but sometimes there is no fault. • Strict Liability doesn’t require Mens Rea only Actus Reus. • Absolute Liability required Mens rea and Actus Reus. • Do we need intention for machine liability? • How do we demonstrate the intentions of a non-human? • How can we do this within existing criminal law principles?
36. Even, Even More Questions… Tort of Negligence 1. Duty of care needs to exist. 2. Breach of that duty of care. 3. That breach of that duty of care cause… 4. …Damage. • Generally, no intention needed. • Conduct is everything!
37. Are The Current International And Domestic Frameworks That Regulate Self Driving Cars Adequate?.
38. A starting point could be the Tort of Negligence.
39. Additional Questions To Consider… • Is General Artificial Intelligence possible? • Will the Singularity/superintelligence occur? • What are the limits of General Artificial Intelligence? • What are the risks General Artificial Intelligence? • What about implications on privacy? • Can machines have consciousness, sentience and mind? • Do/will law prevents and stifle innovation in Artificial Intelligence? • Should different law apply to General Artificial Intelligence and Narrow Artificial Intelligence? • Is the Turing Test is relevant? • Should Artificial Intelligence be given the same legal responsibilities and rights as humans? • Does this include the right to reproduce?
Do you want more like this?
Join below to be notified immediately about new content and more. No annoying daily emails and no spam – just good content when it’s posted.