The researchers at Google DeepMind and LSE developed a text-based game to test if large language models could display behaviors indicative of sentience, such as avoiding pain and seeking pleasure.
The study signifies a pioneering approach to understanding AI capabilities, while emphasizing that current generative AI systems likely do not possess subjective consciousness or sentient characteristics.
Jonathan Birch notes that this research opens a new domain by concluding that we currently lack a thorough test for evaluating sentience in artificial intelligence.
While the study doesn't claim that chatbots are sentient, it provides a starting point for developing future assessments of sentience within complex AI systems.
Collection
[
|
...
]