Apply U Research Team recently delivered a guest talk at the University of Nottingham (UK) at March 2025, exploring the dynamic and evolving world of Large Language Model (LLM) agents. Titled "Large Language Model Agents: Present and Future," the talk brought together researchers, and students to delve deep into the capabilities, current trends, and transformative potential of AI agents.
The talk also introduced how Apply U's own AI-powered systems help students navigate the complex world of university applications and admissions, providing personalized guidance at scale.
Trending in 2025: From Deep Research to Autonomous Agents
The session opened with an overview of 2025's trending technologies in the AI space—from open-source platforms like DeepSeek R1 and CAMEL-AI's OWL to closed systems such as OpenAI O1. Zijian set the stage by clarifying what agents are and how LLM agents differ from basic rule-based or text agents by their ability to reason, plan, and act autonomously.
Climbing the Ladder: From Reasoning to General Intelligence
Zijian explained the different levels of agent sophistication, highlighting how LLM agents progress from basic text generation to complex reasoning tasks. Through relatable examples—like solving logic puzzles or fetching real-time weather data—he demonstrated that LLMs are evolving from simple chatbots into intelligent entities capable of handling nuanced, multi-step problems.
Building Blocks of Smart Agents: Memory, Planning, Tools, and Personas
Attendees were taken through the core capabilities that underpin intelligent agents:
- Memory: How agents remember and learn from interactions
- Planning: The decision-making process behind agent actions
- Tools: The utilities agents use to perform tasks
- Personas: The identity and role of the agent
When One Agent Isn't Enough: Multi-Agent Systems
A standout portion of the talk covered multi-agent systems, where LLMs collaborate, simulate environments, and distribute tasks among themselves.
Opportunities and Challenges: The Human Impact of AI Agents
No exploration of AI would be complete without addressing its real-world consequences. Zijian candidly discussed both the positive impacts—such as personalized education through platforms like Apply U—and the societal challenges, including job displacement, ethical concerns, and digital inequality.