Appendix II. Experiments and Research Initiatives
1. The OpenAI Research: OpenAI conducts extensive research into the safety and ethical implications of artificial general intelligence (AGI). They focus on long-term safety and the development of AI systems that align with human values.
Relevant Work: OpenAI has published papers on the ethical and safety challenges associated with AGI, including how to measure and ensure the alignment of AI systems with human intentions.
2. The AI Alignment Forum:This online community is dedicated to discussing and researching AI alignment, including ethical considerations and indicators of consciousness.
Relevant Work: The forum hosts discussions and research papers on topics such as AI alignment, consciousness, and ethical guidelines for treating AI systems.
3.The Future of Humanity Institute (FHI): Part of the University of Oxford, FHI focuses on global catastrophic risks, including the development of AI. They explore the ethical and existential risks posed by advanced AI systems.
Relevant Work: FHI researchers have published extensively on AI ethics, including the moral and ethical considerations of creating conscious or sentient AI systems.
4.The Machine Intelligence Research Institute (MIRI): MIRI's mission is to ensure that the creation of smarter-than-human intelligence has a positive impact. They focus on the technical and philosophical challenges of AI alignment.
Relevant Work: MIRI has published papers on the theoretical underpinnings of AI alignment and the potential for AI systems to possess consciousness.
5.The Center for Human-Compatible AI (CHAI): Based at UC Berkeley, CHAI focuses on developing AI systems that are provably aligned with human values.
Relevant Work: CHAI researchers investigate the ethical dimensions of AI, including the treatment of AI systems that might exhibit signs of consciousness.
Indicators of Consciousness in AI
Human-like androids, such as Sophia by Hanson Robotics, are not considered conscious by developers despite their advanced capabilities. This is because consciousness involves subjective experiences and self-awareness, which these androids lack. They operate through complex programming and algorithms, mimicking human behavior without genuine feelings or intentionality. Ethical and philosophical considerations also play a role, as attributing consciousness to these machines could lead to misunderstandings and ethical dilemmas. Current AI and robotics technology, while impressive, does not replicate the intricate neural processes of the human brain, reinforcing the distinction between human-like behavior and true consciousness.
However the research into indicators of consciousness in AI is ongoing and multidisciplinary, drawing from neuroscience, cognitive science, and AI. Some key areas of investigation include:
Functional Equivalence: Assessing whether AI systems exhibit behaviors functionally equivalent to conscious beings.
Neuroscientific Approaches: Using insights from neuroscience to identify potential markers of consciousness in AI.
Integrated Information Theory (IIT): Exploring whether AI systems can possess integrated information, a proposed marker of consciousness.
Global Workspace Theory (GWT): Investigating whether AI systems can develop a global workspace, a concept associated with consciousness in humans.