Modelos Dinâmicos Aplicados à Aprendizagem de Valores em Inteligência Artificial

Palavras-chave: Inteligência Artificial, Aprendizagem de Valores, Ciência Cognitiva, Sistemas Dinâmicos.

Resumo

Especialistas em desenvolvimento de Inteligência Artificial (IA) prevêem que o avanço no desenvolvimento de sistemas e agentes inteligentes irá remodelar áreas vitais em nossa sociedade. Contudo, se tal avanço não for realizado de maneira prudente e crítico-reflexiva, pode resultar em desfechos negativos para a humanidade. Por este motivo, diversos pesquisadores na área têm desenvolvido uma concepção de IA robusta, benéfica e segura para a preservação da humanidade e do meio-ambiente. Atualmente, diversos dos problemas em aberto no campo de pesquisa em IA advêm da dificuldade de evitar comportamentos indesejados de agentes e sistemas inteligentes, e ao mesmo tempo especificar o que realmente  queremos que tais sistemas façam, especialmente quando prospectamos a possibilidade de agentes inteligentes atuarem em vários domínios ao longo prazo. É de suma importância que agentes inteligentes artificiais tenham seus valores alinhados com os valores humanos, dado ao fato de que não podemos esperar que uma IA desenvolva valores morais humanos por conta de sua inteligência, conforme é discutido na Tese da Ortogonalidade. Talvez tal dificuldade venha da maneira que estamos abordando o problema de expressar objetivos, valores e metas, utilizando de métodos cognitivos representacionais.  Uma solução para este problema seria a abordagem dinâmica proposta por Dreyfus, que com base na filosofia fenomenológica mostra que a experiência humana do ser-no-mundo em diversos aspectos não é bem representada pelo método cognitivo simbólico ou conexionista, especialmente na questão de aprendizagem de valores. Uma possível abordagem para esse problema seria a utilização de modelos téoricos como SED (situated embodied dynamics) para abordar o porblema de aprendizagem de valores em IA.

Downloads

Não há dados estatísticos.

Biografia do Autor

Nicholas Kluge Corrêa, Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS), Porto Alegre, RS.

Mestre em Engenharia Elétrica pela Pontifícia Universidade Católica do Rio Grande do Sul (Escola Politécnica, PUCRS, Porto Alegre, RS, Brasil) e doutorando em Filosofia (PUCRS) Porto Alegre, RS, Brasil. Bolsista CAPES/PROEX.

Nythamar de Oliveira, Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS), Porto Alegre, RS.

Ph.D. in Philosophy (State University of New York). Professor titular da Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS), Porto Alegre, RS.

Referências

ABADI, M. et al. Deep Learning with Differential Privacy. In: ACM SIGSAC CONFERENCE ON COMPUTER AND COMMUNICATIONS SECURITY (ACM CCS). Proceedings […]. [S. l.]: ACM, 2016. p. 308-318. Disponível em: arxiv.org/abs/1607.00133, 2016. Acesso em: 02 mar. 2020. https://doi.org/10.1145/2976749.2978318.

ALLISON, H. E. Idealism and Freedom: Essays on Kant’s Theoretical and Practical Philosophy. Cambridge: Cambridge University Press, 1996. https://doi.org/10.1017/CBO9781139172875.

AMODEI, D.; OLAH, C.; STEINHARDT, J.; CHRISTIANO, P.; SCHULMAN, J. MANÉ, D. Concrete problems in AI safety. arXiv preprint, [s. l.], 25 July 2016. Disponível em: https://arxiv.org/pdf/1606.06565.pdf. Acesso em: 02 mar. 2020.

ASIMOV, I. I, Robot. New York: Doubleday, 1950.

BARRETT, A. M.; BAUM, S. D. A Model of Pathways to Artificial Superintelligence Catastrophe for Risk and Decision Analysis. Journal of Experimental & Theoretical Artificial Intelligence, JETAI, London, v. 29, n. 2, p. 397 414, 2017. Disponível em: https://arxiv.org/pdf/1607.07730. Acesso em: 02 mar. 2020. https://doi.org/10.1080/0952813X.2016.1186228.

BEER, R. D. Computational and dynamical languages for autonomous agents. In: It’s about time: An overview of the dynamical approach to cognition. Mind as motion: Explorations in the dynamics of cognition. Cambridge, Mass.: MIT Press, 1998. p. 121-147.

BEER, R. D. Dynamical approaches to cognitive science. Trends in Cognitive Sciences. 4. ed. [S. l.: s.n.], 2000. p. 91-99. https://doi.org/10.1016/S1364-6613(99)01440-0.

BEER, R. D. The dynamics of active categorical perception in an evolved model agent (with commentary and response). Adaptive Behavior, Cambridge, Mass, v. 11, n. 4, p. 209-243, dez. 2003. https://doi.org/10.1177/1059712303114001.

BOSTROM, N. The Superintelligent Will: Motivation and Instrumental Rationality in Advance Artificial Agents. Minds and Machines, Dordrecht, Holanda, NL, v. 22, p. 71-85, 2012. https://doi.org/10.1007/s11023-012-9281-3.

BOSTROM, N. Superintelligence. Oxford University Press, 2014. Chapter 12.

BOSTROM, N.; ĆIRKOVIĆ, M. Introduction. In: Bostrom, N.;Ćirković, M. (ed.). Global Catastrophic Risks. New York: Oxford University Press, 2008. p. 1-30.

BRYNJOLFSSON, E.; MCAFEE, A. The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. New York: W.W. Norton Company, 2014.

CHIEL, H. J.; BEER, R. D. The brain has a body: adaptive behavior emerges from interactions of nervous system, body and environment. Trends in neurosciences, Amsterdam, NL, v. 20, n. 12, p. 553-557, 1 Dec. 2012. https://doi.org/10.1016/S0166-2236(97)01149-1.

CHURCHLAND, P. S.; SEJNOWSKI, T. The computational brain. Cambridge, MA: MIT Press, 1992. https://doi.org/10.7551/mitpress/2010.001.0001.

CHURCHILL, R. R.; ULFSTEIN, G. Autonomous Institutional Arrangements in Multilateral Environmental Agreements: A Little-Noticed Phenomenon in International Law. American Journal of International Law, Washington, US, v. 94, n. 4, p. 623-659, 2000. Disponível em: dx.doi.org/10.2307/2589775. Acesso em 02 mar. 2020.

CLARK, A.; CHALMERS, D. J. The extended mind. Analysis, Oxford, v. 58, n. 1, p. 7-19, Jan. 1998. https://doi.org/10.1093/analys/58.1.7.

DEWEY, D. Learning what to value. In: INTERNATIONAL CONFERENCE, AGI 4, August 3–6, 2011, Mountain View, CA, USA. Artificial General Intelligence: proceedings. [S. l: s. n.], 2011. p. 309-314. https://doi.org/10.1007/978-3-642-22887-2_35.

DIJKSTRA, E. W. The threats to computing science. In: ACM SOUTH CENTRAL REGIONAL CONFERENCE, Nov. 16-18, 1984, Austin, TX. Proceedings. [S. l: s. n.], 1984. p. 1-6.

DOCHERTY, B. L. Losing Humanity: The Case Against Killer Robots. New York: Human Rights Watch, 2012.

DREYFUS, H. L. What Computers Still Can’t Do: A critique of Artificial Reason. Cambridge: MIT Press, 1992.

DREYFUS, H. L. Why Heideggerian Artificial Intelligence failed and how fixing it would require making it more Heideggerian. Philosophical Psychology, Abingdon, Inglaterra, v. 20, n. 2, p. 247-268, 2007. https://doi.org/10.1080/09515080701239510.

ELIASMITH, C. The third contender: A critical examination of the dynamicist theory of cognition. Philosophical Psychology, Abingdon, v. 9, n. 4, p. 441-463, 1996. https://doi.org/10.1080/09515089608573194.

FRANKISH, K.; RAMSEY, W. N. The Cambridge handbook of artificial intelligence. Cambridge: Cambridge University Press. 2014. https://doi.org/10.1017/CBO9781139046855.

FREY, C.; OSBORNE, M. The Future of Employment: How Susceptible Are Jobs to Computerisation? Technical Report, Oxford Martin School. Oxford, UK: University of Oxford, 2013.. 6, p. 18-26, 1992.

GÄRDENFORS, P. Conceptual Spaces: The Geometry of Thought. [S. l.]: MIT Press. 2000. https://doi.org/10.7551/mitpress/2076.001.0001.

GIBSON, J. J. The Ecological Approach to Visual Perception. [S. l.]: Houghton Mifflin, 1979.

GRACE, K.; SALVATIER, J.; DAFOE, A.; ZHANG, B.; EVANS, O. When will AI exceed human performance? Evidence from AI experts. arXiv preprint, 3 May 2018. Disponível em: https://arxiv.org/pdf/1705.08807.pdf. Acesso em: 02 mar. 2020.

HADFIELD-MENELL, D.; RUSSELL, S. J.; ABBEEL, P.;DRAGAN, A. Cooperative inverse reinforcement learning. In: ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS NEURAL INFORMATION PROCESSING SYSTEMS CONFERENCE (NIPS), 29., 2016, [S. l.]. Proceedings [S. l.]: NIPS, 2016. p. 3909-3917.

HEIDEGGER, M. Ser e Tempo. Tradução e edição bilíngue com notas de Fausto Castilho. Campinas: Editora Unicamp; Petrópolis: Vozes, 2012. (1927).

HIBBARD, B. The error in my 2001 VisFiles column. [S.l.: s. n.], 2012.

HUME, D. Tratado da Natureza Humana. Tradução de Débora Danowiski. 2.ª ed. São Paulo: Editora da UNESP, 2009. (1739).

JAYNES, E. T. Probability theory: The logic of science. Ed. G. Larry Bretthorst. New York: Cambridge University Press. 2003. Disponível em: doi:10.2277/0521592712. Acesso em: 02 Mar. 2020. https://doi.org/10.1017/CBO9780511790423.

KRIEGESKORTE, N.; KIEVIT, R. A. Representational geometry: integrating cognition, computation, and the brain. Trends in Cognitive Sciences, [s. l.], v.17, n.8, p. 401-12, 2013. https://doi.org/10.1016/j.tics.2013.06.007.

KUZNETSOV, Y. A. Elements of Applied Bifurcation Theory. 3. ed. [S. l.]: Springer. 2004. https://doi.org/10.1007/978-1-4757-3978-7.

LAKOFF, G.; JOHNSON, M. Philosophy in the Flesh. [S.l.]: Basic Books. 1999.

LEGG, S. Is there an elegant universal theory of prediction? In: INTERNATIONAL CONFERENCE ON ALGORITHMIC LEARNING THEORY, ALT 2006,Barcelona, Spain, October 7–10, 2006. Proceedings [...]. [Berlin: Springer], 2006. p. 274-287. https://doi.org/10.1007/11894841_23.

LEIKE, J.; MARTIC, M.; KRAKOVNA, V.; ORTEGA, P.; EVERITT, T.; LEFRANCQ, A.; ORSEAU, L. AI Safety Gridworlds. arXiv preprint, 28 Nov. 2017. Disponível em: https://arxiv.org/pdf/1711.09883.pdf. Acesso em: 03 Mar. 2020.

LLOYD, S. Computational capacity of the universe. Physical Review Letters, New York, v. 88, n. 23, p. 237901, 2002. Disponível em: doi:10. 1103/PhysRevLett.88.237901. Acesso em: 03 Mar. 2020.

MERLEAU-PONTY, M. Phenomenology of Perception. New York: Humanities Press, 1962. (1945).

MORAVEC, H. P. When will computer hardware match the human brain? Journal of Evolution and Technology, [s. l.], v. 1, p. [1-12], 1998. Disponível em: http://www.transhumanist.com/volume1/moravec.htm. Acesso em: 03 Mar. 2020.

MORDVINTSEV, A.; OLAH, C.; TYKA, M. Inceptionism: Going deeper into neural networks. In: Google Research Blog, 17 June 2015. Disponível em: https://ai.googleblog.com/2015/06/inceptionism-going-deeper-intoneural.html. Acesso em: 02 Mar. 2020.

MÜLLER, V. C.; BOSTROM, N. Future progress in artificial intelligence: A survey of expert opinion. In: MÜLLER, V. (ed.). Fundamental issues of artificial intelligence. [S. l.]: Springer, 2016. p. 555-572. https://doi.org/10.1007/978-3-319-26485-1_33.

NEWELL, A. Unified theories of cognition. Cambridge, MA: Harvard University Press, 1990. NG, A. Y.; RUSSELL, S. J. Algorithms for inverse reinforcement learning. In: INTERNATIONAL CONFERENCE ON MACHINE LEARNING (ICML-’00), 17th,[S. l.]. Proceedings. ed. Pat Langley. [S. l: s. n.], 2000. p.663-670.

OMOHUNDRO, S. M. The Nature of Self-Improving Artificial Intelligence. Sept. 5, 2007, revised Jan. 21, 2008. Disponível em: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.137.1199&rep=rep1&-type=pdf. Acesso em: 11 de ago. 2019.

QUINE, W. V. O. Epistemology Naturalized. In: QUINE, W. V. Ontological Relativity and Other Essays. New York: Columbia University Press, 1969. p. 69-90.https://doi.org/10.7312/quin92204-004.

RAIBERT, M. H.; HODGINS, J.K. Biological Neural Networks. Invertebrate Neuroethology and Robotics. [S. l.]: Academic Press, 1993. p. 319-354.

REIMANN, M. W.; NOLTE, M.; SCOLAMIERO, M.;TURNER, K.; PERIN, R.; CHINDEMI, G.; DŁOTKO, P.; LEVI, R.; HESS, K.; MARKRAM, H. Cliques of Neurons Bound into Cavities Provide a Missing Link between Structure and Function. Frontiers in Computational Neuroscience, Lausanne, Switzerland, v. 11, Artigo 48, p. 1-16, 2017. Disponível em: doi: 10.3389/fncom.2017.0004. Acesso em: 02 Mar. 2020. https://doi.org/10.3389/fncom.2017.00048.

ROCKWELL, T. Extended cognition and intrinsic properties. Philosophical Psychology, Abingdon, v. 23, n. 6, p.741-757, 2010. https://doi.org/10.1080/09515089.2010.529044.

ROCKWELL, T. Neither Brain nor Ghost: A Non-Dualist Alternative to the Mind-Brain Identity Theory. Bradford Books: MIT press, 2005. https://doi.org/10.7551/mitpress/4910.001.0001.

SHULMAN, C. Omohundro’s Basic AI Drives and Catastrophic Risks. San Francisco, CA: The Singularity Institute, 2010. Disponível em: https://intelligence.org/files/BasicAIDrives.pdf. Acesso em: 05 Mar. 2020.

SMOLENSKY, P. On the proper treatment of connectionism. Behavioral and Brain Sciences, Cambridge, GB, v. 11, n. 1, p. 1-23, Mar. 1988. https://doi.org/10.1017/S0140525X00052432.

SOARES, N.; FALLENSTEIN, B. Aligning Superintelligence with Human Interests: A Technical Research Agenda. Technical Report. Berkeley, CA: Machine Intelligence Research Institute, 2014.

SOARES, N.; FALLENSTEIN, B.; YUDKOWSKY, E.; ARMSTRONG, S. Corrigibility. In: AAAI WORKSHOPS: WORKSHOPS AT THE TWENTY-NINTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, January 25–26, 2015 Austin, TX., Proceedinds […]. ed. T. Walsh. Palo Alto, CA: AAAI Press, 2015. p. 1-10. (AAAI Technical Report WS-15-02).

SOARES, N. Value Learning Problem. In: ETHICS FOR ARTIFICIAL INTELLIGENCE WORKSHOP AT 25TH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE (IJCAI-2016). New York, NY, USA 9-15 July, 2016. Proceedings […]. [S. l.: s. n.], 2016.

SOTALA, K. From mostly harmless to civilization-threatening: Pathways to dangerous artificial intelligences. In: ECAP10: VIII EUROPEAN CONFERENCE ON COMPUTING AND PHILOSOPHY, 2010, Munich. Proceedings […]. edited by Klaus Mainzer. [S. l.: s. n.], 2010.

SOTALA, K. Disjunctive scenarios of catastrophic AI risk. YAMPOLSKIY, R. V. (ed.). Artificial Intelligence Safety and Security. [S. l.]: Taylor & Francis, 2018. p. 315-337.

SOTALA, K.; YAMPOLSKIY, R. V. Responses to Catastrophic AGI Risk: A Survey. Technical report. Berkeley, CA: Machine Intelligence Research Institute, 2013. https://doi.org/10.1088/0031-8949/90/1/018001.

THAGARD, P. Conceptual revolutions. Princeton: Princeton University Press, 1992. https://doi.org/10.1515/9780691186672.

TVERSKY, A.; Science, New York, v. 211, n. 30, p. 453-458, 30 Jan. 1981. https://doi.org/10.1126/science.7455683.

VAN GELDER, T. The dynamical hypothesis is cognitive science. Behavioral and Brain Sciences. Cambridge, v. 21, n. 5, p. 615- 628, Oct. 1998. https://doi.org/10.1017/S0140525X98001733.

VAN GELDER, T.; PORT, R. It’s about time: An overview of the dynamical approach to cognition. Mind as motion: Explorations in the dynamics of cognition. Cambridge, MA: MIT, 1998.

VON NEUMANN, J.; MORGENSTERN, O. Theory of games and economic behavior. Princeton, NJ: Princeton University Press, 1953.

YUDKOWSKY, E. Artificial Intelligence as a Positive and Negative Factor in Global Risk. In: BOSTROM, N.; ĆIRKOVIĆ, M. M. (ed.). Global Catastrophic Risks, New York: Oxford University Press, 2008. p. 308-45. Disponível em: https://intelligence.org/files/AIPosNegFactor.pdf. Acesso em: 02 Mar 2020.

Publicado
2020-07-27
Como Citar
Corrêa, N. K., & de Oliveira, N. (2020). Modelos Dinâmicos Aplicados à Aprendizagem de Valores em Inteligência Artificial. Veritas (Porto Alegre), 65(2), e37439. https://doi.org/10.15448/1984-6746.2020.2.37439
Seção
Varia