Generative AI + AI Agents + Agentic AI

Let's distinguish between these three types of artificial intelligence (AI):

Generative AI

The AI (a pre-trained transformer, such as ChatGPT or DALL-E) with which the layperson has become most familiar:

• takes in a query with a prompt
• goes through its large language model (LLM), its pre-trained knowledge base
• generates new content (usually text, image, sound or video)

However, its limitations include:

• having a knowledge cutoff (it does not know anything past a certain date)

AI agents

An AI agent refers to a transformer that:

• uses tools (such as some data file, an API or any external data source) to access updated and/or more relevant information
• does all that with very little human intervention (other than an initial prompt)

However, this still has its limitations:

• focuses primarily on prompt-specific output but all not external factors

Agentic AI

An agentic AI then refers to an AI agent system that:

• works on more complex goals
  • i.e. considers more than just a direct information transaction
  • e.g. an AI travel agent booking a flight to another country might consider your ability to go to that country, will probably call an immigration agent
• can work with other agents that can provide more relevant information
  • e.g. an AI travel agent tapping into an airline and hotel agent
• can work on its own but can gather untrained information
  • i.e. could even find tools on its own without prior training