Systems and methods enabling online one-shot learning and generalization by intelligent systems of task-relevant features and transfer to a cohort of intelligent systems

An intelligent system comprising:

• an electronic memory for storing a knowledge representation and storing a first instruction for augmenting the knowledge representation with knowledge involving an object in an environment, the instruction
• provided by a human instructor in a natural language format, and
• representing knowledge related to at least one of an event, an object relation, a rule,
• an action, or a goal;
• one or more subsystems;
• and
• one or more processors coupled to the electronic memory and the one or more subsystems, the one or more processors configured to:
• generate an in-memory representation of a meaning of the instruction;
• divide the in-memory representation of the meaning of the instruction into elements based on predefined knowledge types, where the elements correspond to parts of the in-memory representation of the meaning of the instruction;
• assign to the elements of the in-memory representation of the instruction respective ones of the predefined knowledge types;
• generate one or more executable scripts that encode the at least one of the event, the object relation, the rule, the action, or the goal, where the one or more executable scripts control the one or more subsystems to operate within the environment in order to interact with the object to perform a task, the generate the one or more executable scripts including
• creating at least one of an object specification, an event specification, an object relations specification, an action specification, a rule specification, or a goal specification, based on the assignment of the respective ones of the predefined knowledge types to the elements of the in-memory representation of the instruction, and
• incorporating the at least one of the object specification, the event specification, the object relation specification, the action specification, the rule specification, or the goal specification into the one or more executable scripts;
• and
• augmenting the knowledge representation with knowledge comprising the one or more executable scripts,
• where the generate the in-memory representation of the meaning of the instruction, the divide, the assign, the generate the one or more executable scripts, and the execute steps are performed while the intelligent system is on-line.

A method for operating a robotic system comprising:

• receiving instructions for the robotic system via a natural language interface, including receiving task instructions including a first task instruction;
• and
• interpreting the task instructions accord to data stored in a knowledge repository to form respective operational scripts for controlling the robotic system according to said task instructions, the knowledge repository comprising data representations of knowledge elements, the knowledge elements relating to at least one of events, objects, relationships between objects, actions, and goals, and wherein interpreting the task instructions includes interpreting of the first task instruction by forming an in-memory representation of said instruction according to knowledge elements of the knowledge repository in a first state of said repository, and generating a first script for task the robotic system according to the in-memory representation of the first task instruction;
• wherein receiving the instructions for the robotic system further includes receiving one or more teaching instructions during operation of the robotic system, including receiving a first teaching instruction;
• and
• wherein the method further comprises:
• interpreting the first teaching instruction, including forming an in-memory representation of elements of the first teaching instruction, one or more of the elements of the teaching instruction corresponding to knowledge elements of the knowledge repository in the first state of said repository;
• wherein the interpreting of the first teaching instruction further comprises forming a new knowledge element from the in-memory representation, the new knowledge element representing at least one of an events, object, relationships between objects, actions, rules, and goal represented in terms of one or more knowledge elements of the knowledge repository in the first state of said repository;
• augmenting the knowledge repository with a data representation of the new knowledge element forming a second state of the knowledge repository;
• wherein receiving the task instructions includes receiving a second task instruction after augmenting the knowledge representation to the second state of said repository;
• and
• wherein interpreting the task instructions further comprises interpreting of the second task instruction by forming an in-memory representation of the second task instruction according to knowledge elements of the knowledge repository in the second state of said repository including according to the new knowledge element formed from the first teaching instruction, and generating a second script for operating the robotic system according to the in-memory representation of the second task instruction.