From Search Algorithms to Agentic AI: How Modern AI Systems Are Becoming Smarter
Introduction
Artificial Intelligence is rapidly evolving from simple rule-based programs into intelligent systems that can reason, plan, and make decisions autonomously. In our Artificial Intelligence course, we studied concepts such as AI agents, search algorithms, and problem-solving techniques. Recent research shows how these theoretical ideas are now being applied in modern AI systems.
In this blog, I analyze insights from two recent research papers: “The Rise of Agentic AI: A Review of Definitions, Frameworks, and Challenges (2025)” and “Research on the A Algorithm Based on Adaptive Weights (2025).”* These papers highlight two important directions in AI research: the development of autonomous AI agents and the improvement of search algorithms for better decision making.
Paper 1: The Rise of Agentic AI
Goal of the Paper
The main goal of this paper is to explore the concept of Agentic AI, which refers to AI systems that can operate autonomously to achieve specific goals. Unlike traditional AI systems that only respond to commands, agentic systems can plan tasks, make decisions, and interact with external tools.
The paper reviews different frameworks and architectures used to build AI agents and discusses challenges such as reliability, safety, and evaluation methods.
Key Ideas from the Paper
The research highlights several core components that define agentic AI systems:
- Perception – The ability of an AI system to gather information from its environment.
- Reasoning – Analyzing data and determining possible actions.
- Planning – Creating a sequence of steps to achieve a goal.
- Action – Executing decisions and interacting with external tools or systems.
These components allow AI agents to perform complex tasks such as research assistance, automation, and decision support.
Connection to Our AI Course
This concept strongly connects to the agent models we studied in class. In particular, agentic AI systems resemble goal-based agents, which select actions based on achieving a desired outcome.
However, agentic AI extends this concept further by integrating advanced reasoning, memory, and tool usage. This makes them more capable than simple reflex agents or basic rule-based systems.
Paper 2: Improving the A* Search Algorithm
Goal of the Paper
The second research paper focuses on improving the A* search algorithm, which is one of the most widely used algorithms for pathfinding and problem-solving in artificial intelligence.
Traditional A* evaluates nodes using the function:
f(n) = g(n) + h(n)
Where:
- g(n) represents the cost from the start node to the current node
- h(n) represents the heuristic estimate of the cost from the current node to the goal
The algorithm chooses the node with the lowest total cost.
Problem with the Traditional A* Algorithm
Although A* is efficient and optimal in many cases, it can still expand a large number of nodes in complex search spaces. This can slow down performance in applications such as robotics or navigation systems.
Proposed Improvement
The paper proposes an Adaptive Weighted A* approach that dynamically adjusts the heuristic weight during the search process. The new evaluation function becomes:
f(n) = g(n) + w × h(n)
Here, w is a weight factor that changes during the search.
This allows the algorithm to prioritize faster exploration early in the search while maintaining accuracy later. As a result, the algorithm can reach the goal faster while still producing high-quality solutions.
Connection to Our AI Course
This research directly relates to the search algorithms we studied in class, such as:
- Breadth-First Search (BFS)
- Depth-First Search (DFS)
- Uniform Cost Search (UCS)
- A* Search
Among these algorithms, A* is particularly powerful because it uses heuristics to guide the search toward the goal efficiently. The improved version proposed in the paper shows how researchers continue to refine these algorithms for real-world applications like robot navigation and path planning.
What I Learned from NotebookLM and Manual Reading
After reading the papers manually and exploring them further using NotebookLM, I gained several interesting insights.
First, manual reading helped me understand the core ideas and motivation behind each research paper. It allowed me to see how researchers identify limitations in existing methods and propose improvements.
NotebookLM, on the other hand, helped simplify complex sections and summarize key contributions. It also made it easier to understand technical explanations, particularly in the sections discussing algorithm improvements and agent architectures.
One interesting realization was how modern AI systems combine multiple concepts together. For example, a rescue robot might use:
- AI agents for decision making
- Search algorithms like A* for navigation
- Machine learning models for perception and environment understanding
This integration of multiple AI techniques shows how classroom concepts are applied in real-world systems.
Why These Ideas Matter
The research discussed in these papers represents an important step in the evolution of artificial intelligence.
Agentic AI systems are moving toward autonomous decision-making systems that can handle complex tasks with minimal human intervention. At the same time, improvements in search algorithms such as adaptive A* make problem-solving faster and more efficient.
These advancements will play a significant role in many fields, including:
- Autonomous robotics
- Disaster rescue systems
- Smart assistants
- Navigation and logistics systems
Video Explanation
In addition to this blog post, I created a 2–3 minute video explaining the key ideas from these papers and what I found interesting while studying them. The video provides a short overview of agentic AI and improvements in search algorithms.
(Video link will be embedded here)
Conclusion
Recent developments in AI research demonstrate how foundational concepts like agents and search algorithms are evolving into more advanced and practical systems. Agentic AI represents a shift toward autonomous intelligent systems, while improvements to algorithms like A* enhance the efficiency of problem-solving.
Studying these research papers helped me better understand how theoretical knowledge from our AI course is applied in modern research and real-world applications.
As AI continues to evolve, these technologies will play a crucial role in shaping the next generation of intelligent systems.
Top comments (0)