Alignment Challenges in a Superintelligent AI Landscape
Organisatsiooni nimi
Human Experience & Interaction Design Lab
Kokkuvõte
Research Question: What are the unique challenges to AI alignment in a hypothetical scenario where AI systems surpass human intelligence?
Thesis Outline: Theoretical models: Explore different theoretical models of superintelligence, such as friendly AI and unfriendly AI, to understand potential outcomes.
Value alignment: Investigate how to align a superintelligent AI's values with human values, especially when these values may conflict or be incomplete.
Control problems: Examine the difficulties of maintaining control over a superintelligent AI, including the possibility of it developing its own goals or agency.
Existential risk: Analyze the potential risks posed by a superintelligent AI that is not aligned with human interests, such as causing harm or extinction.
Ethical considerations: Discuss the ethical implications of developing superintelligent AI and the need for responsible development.
Thesis Outline: Theoretical models: Explore different theoretical models of superintelligence, such as friendly AI and unfriendly AI, to understand potential outcomes.
Value alignment: Investigate how to align a superintelligent AI's values with human values, especially when these values may conflict or be incomplete.
Control problems: Examine the difficulties of maintaining control over a superintelligent AI, including the possibility of it developing its own goals or agency.
Existential risk: Analyze the potential risks posed by a superintelligent AI that is not aligned with human interests, such as causing harm or extinction.
Ethical considerations: Discuss the ethical implications of developing superintelligent AI and the need for responsible development.
Lõputöö kaitsmise aasta
2024-2025
Juhendaja
Grace Eden
Suhtlemiskeel(ed)
inglise keel
Nõuded kandideerijale
Tase
Magister
Märksõnad
Kandideerimise kontakt
Nimi
Grace Eden
Tel
E-mail
Vaata lähemalt