意識をもつAIの可能性に関する講演会を実施しました

Aïda Elamrani氏をお招きして、意識をもつAIの可能性に関する講演会を開催しました。

講演者：Aïda Elamrani (Institut Jean Nicod, Ecole Normale Supérieure, PSL University)
講演タイトル：Conscious AI: What Is It and Why Do We Want It?
日時：2024年6月24日（月）
場所：筑波大学　筑波キャンパス春日エリア

講演要旨：
The metaphysical questions surrounding consciousness are widely debated, yet the rapid progress in artificial intelligence (AI) intersects significantly with neuroscience. The computational functions associated with human consciousness are increasingly replicated in artificial agents, raising substantial ethical concerns since the inception of AI. The recent development of Large Language Models, have enabled these agents to formulate plausible ethical judgments, positioning them as explicit ethical agents in machine ethics. As AI progresses toward Artificial Consciousness (AC), it is crucial to evaluate the ethical implications of such agents. Considering the lack of consensus in consciousness studies, we address two scenarios: AC with phenomenal experience (strong AC) and purely functional AC without sentience (weak AC). Our comprehensive literature review and novel arguments indicate that most risks are rooted in fears of artificial suffering and the mishandling of superintelligence, igniting debates over the moral status of artificial agents and their desirability in the global geopolitical and ecological context. However, the potential benefits of AC are more compelling. Research in AC does not necessarily aim to create agents with phenomenal experience. Even if electronic computers could achieve such states, AC technologies do not inherently entail the implementation of valenced states. Moreover, concerns attributed to AC are already present in AI without the consciousness dimension. Crucially, AC systems have the potential to address key ethical issues in current AI, particularly through advanced cognitive functions like empathy and understanding. Lastly, we extend this ethical analysis to the European framework on Trustworthy AI and discuss its global impact. In conclusion, we advocate for a dual meaning of Conscious AI, where artificial consciousness mirrors ethical awareness, aligning advanced cognitive technologies with human values. This vision envisions a future where AI amplifies humanity by caring, not just thinking.

共催：「ヒト脳改変の未来に向けた実験倫理学的ELSI研究方法論の開発」（JPMJRS22J4）（研究代表者：太田紘史