Generative vs Agentic AI: A Comparison
The video provides a description of the key differences and connections between Generative AI (GenAI) and Agentic AI, which are presented as two distinct approaches to artificial intelligence.
Generative AI
Generative AI systems, such as chatbots and image generators, are fundamentally reactive systems. Their core function is to generate some kind of content based upon a specific prompt provided by the user.
Key Characteristics of Generative AI:
• Reactive Nature: GenAI waits for a user to prompt it before taking action.
• Content Generation: It can generate various forms of content, including text, images, code, or audio.
• Mechanism: These systems are essentially sophisticated pattern matching machines. They learn statistical relationships between words, pixels, and waves from massive data sets. When prompted, GenAI predicts what should come next based on its training.
• Limitations: GenAI's work ends at generation; it does not take further steps without the user's input. In applications like content creation, the AI generates possibilities, but a human creator must review, refine, and direct the whole process, curating the generated content.
• Foundation: Chatbots utilize Large Language Models (LLMs) as their backbone, while image and audio generation often use other tools, such as diffusion models.
Agentic AI
Agentic AI systems are proactive systems. While they often begin with a user prompt, that prompt is used to pursue defined goals through a series of actions. Agentic AI thrives in scenarios requiring ongoing management and those consisting of multi-step processes.
Key Characteristics of Agentic AI:
• Proactive Nature: Unlike GenAI, agents decide on and execute actions autonomously to achieve a goal.
• Life Cycle: An agentic system generally follows a life cycle: it first perceives its environment, then decides an action to take, executes that action, and then learns from the output, often completing this loop with minimal human intervention.