Žalos sukeltos naudojant dirbtinį intelektą teisinio reguliavimo problematika Lietuvos dirbtinio intelekto strategijos kontekste
Lėckaitė, Deimantė |
Dirbtinis intelektas gali turėti naudingų ir žalingų padarinių, atsižvelgiant į dabartinius jo raidos tempus. Dėl šios priežasties būtina nustatyti gaires dabartinio ir tolesnio dirbtinio intelekto plėtojimui ir panaudojimui, etinių ir teisinių problemų, susijusių su dirbtinio intlekto galimu šališku sprendimų priėmimu, bei atsakomybės klausimais. Dabartinis teisinis reguliavimas Lietuvoje nenumato teisinio aiškumo dėl atsakomybės už žalos, sukeltos dirbtinio intelekto, atlyginimo, kas sąlygojo šio magistrinio darbo temos pasirinkimą, siekiant išsiaiškinti, ar Lietuvos dirbtinio intelekto strategijoje yra tinkamai įvertinti teisinis reguliavimas ir numatytos galimos rizikos, naudojant dirbtinį intelektą, atsižvelgiant į naujus iššūkius. Šiuo magistriniu darbu siekiama, apibendrinus skirtingų šalių galimus dirbtinio intelekto strategijos sukeltai žalai atlyginti teisinio reguliavimo metodus, nustatyti efektyviausius dirbtinio intelekto sukeltos žalos atlyginimo teisinio reguliavimo instrumentus. Pirmame šio darbo skyriuje nustatyta, jog robotai šiuo metu jau geba savarankiškai priimti sprendimus, todėl kyla poreikis sukurti tinkamai pritaikytą taisyklių sistemą, siekiant suvaldyti rizikos faktorius, susijusius su sistemų veikimo paaiškinamumu, numatomumu ir atsekamumu, veikiant tokiems dirbtinio intelekto metodams, kaip gilusis mokymasis, todėl tolesnei dirbtinio intelekto plėtrai reikalinga pakankamai lanksti reglamentavimo sistema, skatinanti inovacijas ir kartu užtikrinanti aukštą apsaugos ir saugos lygį. Antrame skyriuje išnagrinėtas Japonijos ir Prancūzijos dirbtinio intelekto strategijas palyginus su Lietuvos dirbtinio intelekto strategija, padaryta išvada, kad Lietuvoje dirbtinio intelekto technologijų ir autonominių sistemų atsakingo naudojimo klausimais vis dar egzistuoja etikos ir teisės problemų. Valstybė turėtų spartinti mokslinius tyrimus, rengti daugiau specialistų, diegti šias technologijas valstybės institucijose, viešosiose paslaugose, siekiant efektyvinti elektroninę demokratiją, stiprinti teisinę bazę, bei didinti visuomenės pasitikėjimą šiomis technologijomis. Trečiame skyriuje nustatyti, dirbtinio intelekto sistemos sukeltos žalos, galimos atsakomybės paskirstymo, sprendimo būdai yra pagrįsti, tačiau numanomas tam tikras jų tobulinimas ir koregavimas. Nors atitinkamai Rezoliucijoje, Direktyvoje, bei moksliniuose straipsniuose pateikiami atsakingo už roboto sukeltą žalą subjekto identifikavimo būdai, tačiau manytina, kad gali būti problematiška nustatyti, kiek kiekvienas subjektas, kūręs dirbtinį intelektą, gali būti atsakingas už padarytą žalą. Įvertinus galimas civilinės atsakomybės pritaikymo rūšis (metodus) bei jų taikyme galimus privalumus ir trūkumus, nustačius saugumo normas ir civilinę atsakomybę reglamentuojančią saugią teisinę bazę dirbtinio intelekto technologijoms, pasiektas magistrinio darbo tikslas, bei parengtos rekomendacijos dėl dirbtinio intelekto sukeltos žalos atlyginimo teisinio reguliavimo Lietuvoje.
Artificial intelligence can bring both positive and negative effects, since taking into consideration the speed of its current development, this technology can become independent and superintelectual in a few decades. Therefore, guidelines must be set for current and future development and use of artificial intelligence to solve ethical and legal issues related to possible bias decision-making of artificial intelligence and its liability. To reach this aim every member-state is encouraged to draft an artificial intelligence strategy, which would regulate the reimbursement of the damages caused by artificial intelligence.The fact that the current legal regulation in Lithuania does not provide any legal clarification with regard to the liability for the damage caused by artificial intelligence influenced the selection of the topic for this master thesis, which attempts to determine whether the possible risks and legal regulation of using artificial intelligence set in the strategy of artificial intelligence of Lithuania are adequately assessed taking into consideration the most recent challenges in this field. Therefore, having summarised the strategies ofthe reimbursement of the possible damages caused by artificial intelligence and legal regulation methods employed by other countries, this master thesis aims at determining the most effective legal regulation tools for the reimbursement of damages caused by artificial intelligence which would help with the implementation of the Lithuanian artificial intelligence strategy. Chapter 1 of this thesis establishes that the continuous advancement of technologies, which, nowadays, allows robots to make decisions independently, requires creating a well-adapted system of rules to control such risk factors as certain level of unpredictability of robot behaviour, unique and unpredictable robot interaction with the environment and essential problemsof clarification, predictability and traceability of system functioning, when such artificial intelligence methods as deep learning are used. Furthermore, it has been determined that if the harmful behaviour of robot could have been forseen, the legal liability for the damages caused by robot’s harm is assumed by a certain human subject, for instance a manufacturer, operator, owner or user. Further development of artificial intelligence requires sufficiently flexible regulation system which encourages innovations and, at the same time, ensures high level of safety and security. Chapter 2 deals with the artificial intelligence strategies in Japan and France and compares them to the Lithuanian strategy of artificial intelligence. The main disadvantages of the strategy are that the country provides little to no encouragement for the development of these technologies, there is a lack of governmental programmes related to these technologies, business investments are low, therefore a vacuum of ethics and law related to the issues of responsible use of artificial intelligence and autonomous systems still exists in Lithuania. Due to this fact and following the example of foreign countries, the Country should intensify scholarly research, educate more specialists, implement these technologies in state institutions and public service sector, make electronic democracy more effective, strengthen legal basis and increase public trust in new technologies. There is no universally applied legal regulation of artificial intelligence, however it is important to create such regulation which would be based on human rights and consider human welfare as its main goal. Successful implementation of artificial intelligence strategy would allow Lithuania to not only improve the conditions for innovative businesses and citizens of the country, but also contribute to the attempts of the EU to compete with other leading countries that conduct active research and innovations in the field of artificial intelligence. This goal will be reached only if Lithuania puts into effect the set methods of guideline implementation which, based on the comparative analysis, would be equivalent to the artificial intelligence strategies of Japan and France. Chapter 3 sets forth that if the damage is caused when the driver is personally operating the autonomous vehicle, current legal regulation should be applied, however if the driver fails to take control of the vehicle on time, the damage incurred by the third parties should be reimbursed by both the driver and the manufacturer of the autonomous vehicle. Having analysed the legal acts which regulate the solving of the issue of liability arising from the actions of a smart robot which were not foreseen by its developer, it is assumed that although valid rules provide certain guidelines, the victims find it hard to prove the causal relation between the damage and the defect of the product. Due to the characteristics of smart robots, strict liability and risk management method, which could be implemented by establishing special remuneration fund, should be applied in certain cases. Generally, strict liability should be applied to the subject who taught the smart robot certain actions (if such teacher can be identified), while risk management method should be applied in cases when a smart robot learned the actions that caused damage independently. Compulsory insurance of the most advanced smart robots, which would be applied not to the manufacturer but to the operator that in the majority of cases would bethe same person as the teacher of the smart robot, should be established accordingly. Similarly to the case of motor vehicle insurance, it is offered to complement such insurance system with a fund, established to ensure the reimbursement of damages when the insurance is not present. Having evaluated the possibilities of the electronic person to be considered a subject of legal liability, the conclusion can be drawn that the awarding of legal capacity to robots would be an unsuitable legal and ethical perspective which would cause more harm than good, since it would provide the manufacturers and the developers with an opportunity to avoid liability.Therefore, the methods of solution of damage caused by artificial intelligence and possible awarding of liability are grounded, however certain improvements and corrections can be made. Thus, problems may arise when establishing the contribution of every subject that developed the artificial intelligence to the liability. Having considered possible application types (methods) of civil liability as well as advantages and disadvantages of their application and setting a safe legal base regulating the safety norms and civil liability of artificial intelligence, the aim of the master thesis has been reached and the recommendations for the legal regulation of the reimbursement of damage caused by artificial intelligence in Lithuania have been provided.