Human-like AI raises deception risks for users

Significance

The perception of AI sentience as real by an individual such as Lemoine -- with full awareness of the synthetic nature of LaMDA and technical knowledge of its workings -- questions whether a belief in AI sentience should be dismissed as indisputably erroneous.

Impacts

• Current regulations are insufficient to mitigate the potential effects of human-like AI in different applications.
• Policymakers may need to plan measures to avoid confusion between an AI able to imitate human sentient experience and actual AI sentience.
• There is insufficient understanding of the impact on users of interactions with AI experientially indistinguishable from humans.