Revisión sistemática de taxonomías de riesgos asociados a la Inteligencia Artificial

Contenido principal del artículo

Guillem Bas Graells
Roberto Tinoco Devia
https://orcid.org/0009-0004-7763-6979
Claudette Salinas Leyva
Jaime Sevilla Molina
https://orcid.org/0000-0002-4454-1146

Resumen

Este artículo realiza una revisión sistemática de treinta y seis taxonomías de riesgos asociados a la Inteligencia Artificial (IA) que se han realizado desde el 2010 hasta la fecha, utilizando como metodología el protocolo Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). El estudio se basa en la importancia de estas para estructurar la investigación de los riesgos y para distinguir y definir amenazas. Ello permite identificar las cuestiones que generan mayor preocupación y, por lo tanto, requieren mejor gobernanza. La investigación permite extraer tres conclusiones. En primer lugar, se observa que la mayoría de los estudios se centran en amenazas como la privacidad y la desinformación, posiblemente debido a su concreción y evidencia empírica existente. Por el contrario, amenazas como los ciberataques y el desarrollo de tecnologías estratégicas son menos citadas, a pesar de su creciente relevancia. En segundo lugar, encontramos que los artículos enfocados en el origen del riesgo tienden a considerar más frecuentemente riesgos extremos en comparación con los trabajos que abordan las consecuencias. Esto sugiere que la literatura ha sabido identificar las potenciales causas de una catástrofe, pero no las formas concretas en las que esta se puede materializar en la práctica. Finalmente, existe una cierta división entre aquellos artículos que tratan daños tangibles presentes y aquellos que cubren daños potenciales futuros. No obstante, varias amenazas se tratan en la mayoría de los artículos de todo el espectro indicando que existen puntos de unión entre clústeres.

Palabras clave:
Inteligencia artificial Riesgo Amenaza Daño Perjuicio Taxonomía

Citas

Acemoglu, D. (2021). Harms of AI (Working Paper 29247). National Bureau of Economic Research. https://doi.org/10.3386/w29247

Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610-623. https://doi.org/10.1145/3442188.3445922

Blauth, T. F., Gstrein, O. J., & Zwitter, A. (2022). Artificial intelligence crime: An overview of malicious use and abuse of AI. IEEE Access, 10, 77110-77122.

https://ieeexplore.ieee.org/document/9831441

Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., ... & Liang, P. (2021). On the opportunities and risks of foundation models. arXiv preprint arXiv: 2108.07258. https://doi.org/10.48550/arXiv.2108.07258

Brundage, M., Avin, S., Clark, J., Toner, H., Eckersley, P., Garfinkel, B., ... & Amodei, D. (2018). The malicious use of artificial intelligence: Forecasting, prevention, and mitigation. arXiv preprint arXiv:1802.07228. https://doi.org/10.48550/arXiv.1802.07228

Buçinca, Z., Pham, C. M., Jakesch, M., Ribeiro, M. T., Olteanu, A., & Amershi, S. (2023). AHA!: Facilitating AI Impact Assessment by Generating Examples of Harms. arXiv preprint arXiv:2306.03280. https://doi.org/10.48550/arXiv.2306.03280

Bucknall, B. S., & Dori-Hacohen, S. (2022). Current and near-term AI as a potential existential risk factor. In Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society (pp. 119-129). https://doi.org/10.1145/3514094.3534146

CAIS (2023). Statement on AI Risk. Center of AI Safety. https://www.safe.ai/statement-on-ai-risk

Chan, A., Salganik, R., Markelius, A., Pang, C., Rajkumar, N., Krasheninnikov, D., Langosco, L., He, Z., Duan, Y., Carroll, M., Lin, M., Mayhew, A., Collins, K., Molamohammadi, M., Burden, J., Zhao, W., Rismani, S., Voudouris, K., Bhatt, U., … Maharaj, T. (2023). Harms from Increasingly Agentic Algorithmic Systems. In 2023 ACM Conference on Fairness, Accountability, and Transparency (pp.0 651-666). https://doi.org/10.1145/3593013.3594033

Clarke, S. & Whittlestone, J. (2022). A Survey of the Potential Long-term Impacts of AI. In Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society (pp. 192-202). https://doi.org/10.1145/3514094.3534131

Critch, A., & Russell, S. (2023). TASRA: A Taxonomy and Analysis of Societal-Scale Risks from AI. arXiv preprint arXiv:2306.06924. https://doi.org/10.48550/arXiv.2306.06924

Ganguli, D., Lovitt, L., Kernion, J., Askell, A., Bai, Y., Kadavath, S., ... & Clark, J. (2022). Red teaming language models to reduce harms: Methods, scaling behaviors, and lessons learned. arXiv preprint arXiv:2209.07858. https://n9.cl/u57rh

Garvey, C. (2018). AI Risk Mitigation Through Democratic Governance: Introducing the 7-Dimensional AI Risk Horizon. In Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society (pp. 366-367). https://doi.org/10.1145/3278721.3278801

Guan, H., Dong, L., & Zhao, A. (2022). Ethical Risk Factors and Mechanisms in Artificial Intelligence Decision Making. Behavioral Sciences, 12(9), Article 9. https://doi.org/10.3390/bs12090343

Guembe, B., Azeta, A., Misra, S., Osamor, V. C., Fernandez-Sanz, L., & Pospelova, V. (2022). The Emerging Threat of Ai-driven Cyber Attacks: A Review. Applied Artificial Intelligence, 36(1), 2037254. https://doi.org/10.1080/08839514.2022.2037254

Hagerty, A. & Rubinov, I. (2019). Global AI ethics: a review of the social impacts and ethical implications of artificial intelligence. arXiv preprint arXiv:1907.07892. https://doi.org/10.48550/arXiv.1907.07892

Hazell, J. (2023). Large language models can be used to effectively scale spear phishing campaigns. arXiv preprint arXiv:2305.06972. https://n9.cl/siuce

Hendrycks, D., Mazeika, M., & Woodside, T. (2023). An Overview of Catastrophic AI Risks. arXiv preprint arXiv:2306.12001. https://doi.org/10.48550/arXiv.2306.12001

Kilian, K. A., Ventura, C. J., & Bailey, M. M. (2023). Examining the Differential Risk from High-level Artificial Intelligence and the Question of Control. Futures, 151, 103182. https://doi.org/10.1016/j.futures.2023.103182

Kirk, H. R., Vidgen, B., Röttger, P., & Hale, S. A. (2023). Personalisation within bounds: A risk taxonomy and policy framework for the alignment of large language models with personalised feedback. arXiv preprint arXiv:2303.05453. https://doi.org/10.48550/arXiv.2303.05453

Kuleshov, A., Ignatiev, A., & Abramova, A. (2021). The Deficiency of “Redline/Greenline” Approach to Risk Management in AI Applications. In 2021 International Conference Engineering Technologies and Computer Science (EnT) (pp. 49-55). https://doi.org/10.1109/EnT52731.2021.00015

Manheim, Karl M. and Kaplan, Lyric, Artificial Intelligence: Risks to Privacy and Democracy (October 25, 2018). 21 Yale Journal of Law and Technology 106 (2019), Loyola Law School, Los Angeles Legal Studies Research Paper No. 2018-37, Available at SSRN: https://ssrn.com/abstract=3273016

McLean, S., Gemma J. M. Read, Jason Thompson, Chris Baber, Neville A. Stanton & Paul M. Salmon (2023) The risks associated with Artificial General Intelligence: A systematic review. Journal of Experimental & Theoretical Artificial Intelligence, 35(5), 649-663. https://doi.org/10.1080/0952813X.2021.1964003

Meek, T., Barham, H., Beltaif, N., Kaadoor, A., & Akhter, T. (2016). Managing the ethical and risk implications of rapid advances in artificial intelligence: A literature review. In 2016 Portland International Conference on Management of Engineering and Technology (PICMET) (pp. 682-693). https://doi.org/10.1109/PICMET.2016.7806752

Müller, V. C. (2020). Ethics of Artificial Intelligence and Robotics. En E. Zalta (Ed.), Stanford Encyclopedia of Philosophy (pp. 1-70). CSLI, Stanford University.

O’Neill, M. & Connor, M. (2023). Amplifying Limitations, Harms and Risks of Large Language Models. arXiv. (arXiv:2307.04821). https://doi.org/10.48550/arXiv.2307.04821

OpenAI (2023, marzo 15). GPT-4 Technical Report. arXiv.Org. https://n9.cl/zfb8z

Page, J., Bain, M., & Mukhlish, F. (2018, August). The risks of low level narrow artificial intelligence. In 2018 IEEE international conference on intelligence and safety for robotics (ISR) (pp. 1-6). https://ieeexplore.ieee.org/document/8535903/

Page, M. J., McKenzie, J. E., Bossuyt, P. M., Boutron, I., Hoffmann, T. C., Mulrow, C. D., Shamseer, L., Tetzlaff, J. M., Akl, E. A., Brennan, S. E., Chou, R., Glanville, J., Grimshaw, J. M., Hróbjartsson, A., Lalu, M. M., Li, T., Loder, E. W., Mayo-Wilson, E., McDonald, S., … Moher, D. (2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. Systematic Reviews, 10(1), 89. https://n9.cl/hgnin

Rajendra, P., Kumari, M., Rani, S., Dogra, N., Boadh, R., Kumar, A., & Dahiya, M. (2022). Impact of artificial intelligence on civilization: Future perspectives. Materials Today: Proceedings, 56, 252-256. https://doi.org/10.1016/j.matpr.2022.01.113

Scherer, M. U. (2015). Regulating Artificial Intelligence Systems: Risks, Challenges, Competencies, and Strategies (SSRN Scholarly Paper 2609777). https://n9.cl/muwp6

Shelby, R., Rismani, S., Henne, K., Moon, A., Rostamzadeh, N., Nicholas, P., ... & Virk, G. (2023). Sociotechnical harms of algorithmic systems: Scoping a taxonomy for harm reduction. In Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society (pp. 723-741). https://doi.org/10.48550/arXiv.2210.05791

Soice, E. H., Rocha, R., Cordova, K., Specter, M., & Esvelt, K. M. (2023). Can large language models democratize access to dual-use biotechnology? arXiv preprint ar- Xiv:2306.03809. https://arxiv.org/abs/2306.03809v1

Solaiman, I., Talat, Z., Agnew, W., Ahmad, L., Baker, D., Blodgett, S. L., ... & Vassilev, A. (2023). Evaluating the Social Impact of Generative AI Systems in Systems and Society. arXiv preprint arXiv:2306.05949. https://arxiv.org/abs/2306.05949v2

Strasser, A. (2023). On pitfalls (and advantages) of sophisticated large language models. arXiv preprint arXiv:2303.17511. https://doi.org/10.48550/arXiv.2303.17511

Tamkin, A., Brundage, M., Clark, J., & Ganguli, D. (2021). Understanding the capabilities, limitations, and societal impact of large language models. arXiv preprint ar-Xiv:2102.02503. https://doi.org/10.48550/arXiv.2102.02503

Turchin, A. & Denkenberger, D. (2020). Classification of global catastrophic risks connected with artificial intelligence. AI & Society, 35(1), 147-163. https://n9.cl/uob2z

Urbina, F., Lentzos, F., Invernizzi, C., & Ekins, S. (2022). Dual use of artificial-intelligence- powered drug discovery. Nature Machine Intelligence, 4(3), Article 3. https://doi.org/10.1038/s42256-022-00465-9

Vesnic-Alujevic, L., Nascimento, S., & Pólvora, A. (2020). Societal and ethical impacts of artificial intelligence: Critical notes on European policy frameworks. Telecommunications Policy, 44(6), 101961. https://doi.org/10.1016/j.telpol.2020.101961

Vold, K., & Harris, D. R. (2021). How Does Artificial Intelligence Pose an Existential Risk? En C. Véliz (Ed.), Oxford Handbook of Digital Ethics. Oxford University Press.

Weidinger, L., Mellor, J., Rauh, M., Griffin, C., Uesato, J., Huang, P. S., ... & Gabriel, I. (2021). Ethical and social risks of harm from language models. arXiv preprint ar- Xiv:2112.04359. https://doi.org/10.48550/arXiv.2112.04359

Yampolskiy, R. V. (2015). Taxonomy of Pathways to Dangerous Artificial Intelligence. (arXiv:1511.03246). arXiv. https://doi.org/10.48550/arXiv.1511.03246

Detalles del artículo