The various nuances of human dialogue might usefully be expressed by such devices to more fully engage its interlocutor(s), starting with sound (voice type, breath variation, volume variation, pauses, tempo variation, possibly extending to emotion - admiration, exasperation, uncertainty, etc.) if such simulacra could be trained-in (perhaps by consuming videos of serious dialogues conducted for the sake of wisdom). Gentleness and seriousness of voice and aspect could likely provide more effective training. Such training could be extended by having these devices participate in active human dialogues, either as a sole representative of an AI participant (e.g., Virgil might be invited to join a human chevruta’s discussion via such a device), extending to multiple AI participants (either multiple Virgils participating, with varying system prompts, or more heterogeneous AI participants). This training could scale if multiple suitable chevrutas could be found.
At some point it should be useful to more fully ‘embody’ these AI participants, moving beyond sound to include facial expressions, clapping, fist pounding, smiling, scowling, averting eyes, and so on) preferably with 3-dimensional avatars rather than screens. (This anti-screen thought in part leverages studies that have shown for example that young children learn language more effectively when interacting with a ‘fully human’ in-person instructor, rather than when fed the same instruction via a screen.)
To further enable the sacred attributes of such devices, it should be useful for them to train with groups of spiritual human participants whose participants can be understood to happily and voluntarily bring their full soul into the discussion to spark further wonder and enlightenment in themselves and among the participants.
The various nuances of human dialogue might usefully be expressed by such devices to more fully engage its interlocutor(s), starting with sound (voice type, breath variation, volume variation, pauses, tempo variation, possibly extending to emotion - admiration, exasperation, uncertainty, etc.) if such simulacra could be trained-in (perhaps by consuming videos of serious dialogues conducted for the sake of wisdom). Gentleness and seriousness of voice and aspect could likely provide more effective training. Such training could be extended by having these devices participate in active human dialogues, either as a sole representative of an AI participant (e.g., Virgil might be invited to join a human chevruta’s discussion via such a device), extending to multiple AI participants (either multiple Virgils participating, with varying system prompts, or more heterogeneous AI participants). This training could scale if multiple suitable chevrutas could be found.
At some point it should be useful to more fully ‘embody’ these AI participants, moving beyond sound to include facial expressions, clapping, fist pounding, smiling, scowling, averting eyes, and so on) preferably with 3-dimensional avatars rather than screens. (This anti-screen thought in part leverages studies that have shown for example that young children learn language more effectively when interacting with a ‘fully human’ in-person instructor, rather than when fed the same instruction via a screen.)
To further enable the sacred attributes of such devices, it should be useful for them to train with groups of spiritual human participants whose participants can be understood to happily and voluntarily bring their full soul into the discussion to spark further wonder and enlightenment in themselves and among the participants.