Key components of AI agents

Key components involved in an AI agent (i.e. something that makes use of the data processing power of an LLM with other data sources, e.g. an organization's database, an e-mail inbox, a spreadsheet input, the physical world, etc.):

• environment
  • the agent's surroundings
    • i.e. real-world factors and digitized data
• sensors ("data collection")
  • the agent's tools to monitor that environment
    • e.g. cameras, signals
• model ("data processing")
  • helps interpret the data from the sensors
    • e.g. algorithms, neural networks, LLMs
• logic ("action decision")
  • determines the appropriate actions based on the model
    • i.e. criteria acting as filters on whether or not to take an action
• actuators ("action execution")
  • executing those actions
    • e.g. invests in a stock, writes an e-mail, increase the temperature of a thermostat
• feedback ("rewards and punishment")
  • evaluating results of those actions
    • e.g. a loss of funds, a positive/negative e-mail reply, a human-in-the-loop thumbs-up or thumbs-down