Human - AI Agents

Human - AI Agents is a collaboration form that involve at least a single human and one or more AI Agents. In this context of interaction, we have the Human be the ultimate initiator of the collaboration, in which the human assign mission and tasks to the AI Agents. Mission will be something that comprise of more than a single task, while task are one-off thing for the AI Agents to do. That being said, the task here is not necessarily be a short-lived one, since it may also be a repetitive thing to be done over certain period of time. Nevertheless task is always predefined. For instance "Categorize the dataset obtained into data that reinforce hypothesis and data that question the hypothesis." is a task.

Mission, however, is not as well-defined as task. Something like "Search and Discover a set of alternative that improve material properties to >5% of existing benchmark." is considered a mission.

The Human - AI Agents interaction is what we commonly refer as human-in-the-loop approach in development of either machine learning protocol, or any other artificial intelligence pipeline. Here in the context of AI4DeSci, the collaboration can happen either with a single AI Agents, or with many AI Agents, that could be either copy of each others or each distinct AI Agents.