Customization of a neural-network persona
A computer system that customizes an output provided by a pretrained neural network is described. During operation, the computer system may receive a prompt (or input) associated with a first individual. Then, the computer system may access stored information (in memory) associated with a persona associated with a second individual. For example, the second individual may be different from the first individual. However, in some embodiments, the second individual may be the same as the first individual. Note that the stored information may include memories associated with one or more different sessions of interacting with the pretrained neural network than a current session. Moreover, the computer system may modify the prompt based at least in part on the persona associated with the second individual. Next, the computer system may generate, using the pretrained neural network, the output based at least in part on the modified prompt.
1 . A computer system, comprising:
a computation device;
memory configured to store program instructions, wherein, when executed by the computation device, the program instructions cause the computer system to perform operations comprising:
receiving a prompt associated with a first individual;
accessing, in the memory, stored information associated with a persona associated with a second individual;
modifying the prompt based at least in part on the persona associated with the second individual, wherein the modifying comprises computing an embedding that is input to the pretrained neural network, and wherein the computing comprises:
determining match scores between the prompt and different portions of the information;
ranking the different portions based at least in part on the match scores; and
selecting at least the part of the persona associated with the second individual based at least in part on the ranking; and
generating, using a pretrained neural network, an output based at least in part on the modified prompt.
2 . The computer system of claim 1 , wherein the second individual is different from the first individual.
3 . The computer system of claim 1 , wherein the stored information comprises memories associated with one or more different sessions of interacting with the pretrained neural network than a current session.
4 . The computer system of claim 3 , wherein the one or more different sessions of interaction are with the first individual, the second individual or one or more different individuals.
5 . The computer system of claim 1 , wherein the information customizes one or more of: an appearance of a dynamic virtual representation provided by the pretrained neural network, a communication style of the dynamic virtual representation provided by the pretrained neural network, a personality of the dynamic virtual representation provided by the pretrained neural network, motivation or goals of the dynamic virtual representation provided by the pretrained neural network, or knowledge of the dynamic virtual representation provided by the pretrained neural network.
6 . The computer system of claim 1 , wherein the output comprises: text, audio, a song, and/or an image.
7 . The computer system of claim 1 , wherein the pretrained neural network comprises a large language model (LLM).
8 . The computer system of claim 1 , wherein the operations comprises: receiving, associated with the second individual, at least some of the information; receiving, associated with the second individual, selections of at least some of the information from predefined lists of information; or both.
9 . The computer system of claim 1 , wherein the modified prompt, the output or both are generated based at least in part on one or more predefined templates.
10 . The computer system of claim 9 , wherein the one or more predefined templates comprise a predefined response to a type of prompt.
11 . The computer system of claim 1 , wherein the modified prompt comprises: text, metadata associated with the prompt, one or more variables, and one or more actions.
12 . The computer system of claim 1 , wherein the persona corresponds to a dynamic virtual representation associated with the second individual.
13 . A non-transitory computer-readable storage medium for use in conjunction with a computer system, the computer-readable storage medium configured to store program instructions that, when executed by the computer system, causes the computer system to perform operations comprising:
receiving a prompt associated with a first individual;
accessing, in memory, stored information associated with a persona associated with a second individual;
modifying the prompt based at least in part on the persona associated with the second individual, wherein the modifying comprises computing an embedding that is input to the pretrained neural network, and wherein the computing comprises:
determining match scores between the prompt and different portions of the information;
ranking the different portions based at least in part on the match scores; and
selecting at least the part of the persona associated with the second individual based at least in part on the ranking; and
generating, using a pretrained neural network, an output based at least in part on the modified prompt.
14 . The non-transitory computer-readable storage medium of claim 13 , wherein the stored information comprises memories associated with one or more different sessions of interacting with the pretrained neural network than a current session; and
wherein the one or more different sessions of interaction are with the first individual, the second individual or one or more different individuals.
15 . The non-transitory computer-readable storage medium of claim 13 , wherein the information customizes one or more of: an appearance of a dynamic virtual representation provided by the pretrained neural network, a communication style of the dynamic virtual representation provided by the pretrained neural network, a personality of the dynamic virtual representation provided by the pretrained neural network, motivation or goals of the dynamic virtual representation provided by the pretrained neural network, or knowledge of the dynamic virtual representation provided by the pretrained neural network.
16 . A method for customizing an output provided by a pretrained neural network, comprising:
by a computer system:
receiving a prompt associated with a first individual;
accessing, in memory, stored information associated with a persona associated with a second individual;
modifying the prompt based at least in part on the persona associated with the second individual, wherein the modifying comprises computing an embedding that is input to the pretrained neural network, and wherein the computing comprises:
determining match scores between the prompt and different portions of the information;
ranking the different portions based at least in part on the match scores; and
selecting at least the part of the persona associated with the second individual based at least in part on the ranking; and
generating, using the pretrained neural network, the output based at least in part on the modified prompt.
17 . The method of claim 16 , wherein the stored information comprises memories associated with one or more different sessions of interacting with the pretrained neural network than a current session; and
wherein the one or more different sessions of interaction are with the first individual, the second individual or one or more different individuals.
18 . The method of claim 16 , wherein the information customizes one or more of: an appearance of a dynamic virtual representation provided by the pretrained neural network, a communication style of the dynamic virtual representation provided by the pretrained neural network, a personality of the dynamic virtual representation provided by the pretrained neural network, motivation or goals of the dynamic virtual representation provided by the pretrained neural network, or knowledge of the dynamic virtual representation provided by the pretrained neural network.
19 . The method of claim 16 , wherein the second individual is different from the first individual.
20 . The method of claim 16 , wherein the method comprises: receiving, associated with the second individual, at least some of the information; receiving, associated with the second individual, selections of at least some of the information from predefined lists of information; or both.