May 2025 AI & Tech Highlights: Agents, VR Education, and Data Harmonization

AI Agents vs. Generative AI

• 🤖 AI agents represent the next step beyond generative AI, focusing on the digitization of specific skills and autonomous operation.
• 🔑 Unlike generative AI tools that unlock proprietary data, agents are designed to distribute work and achieve objectives without direct human intervention.
• 💡 The current generation of agents digitizes narrow skills, akin to self-driving cars with specific functions, rather than aiming for Artificial General Intelligence (AGI).
• 🧠 Agentic AI systems have a more complex architecture than LLM-based generative AI, including a knowledge graph, memory, and interfaces for real-world interaction (tool use).

Polars for Data Processing Efficiency

• ⚡ The utility company Aliander significantly improved data processing by migrating from Python/Pandas and R to Polars.
• 📊 This transition reduced memory usage from 700 GB to 40 GB for a 500 GB job, drastically increasing efficiency and lowering costs.
• 📚 The development of the O'Reilly book on Polars and its real-world implementation at Aliander were intertwined, revealing limitations and driving improvements in the library.

VR in Education and Training

• 🎓 Virtual Reality (VR) platforms like SEEK offer a solution to teacher shortages by enabling a single instructor to teach at scale.
• 🚀 VR is crucial for training in new industries, such as autonomous vehicles, allowing for safe and cost-effective training of large numbers of individuals (e.g., 100,000 pilots).
• 🧠 VR creates immersive experiences that build memory and practical experience, enabling individuals to gain skills (like flying an aircraft or performing CPR) before engaging with real-world scenarios.
• 🏥 VR training is particularly impactful in healthcare for skills like CPR and intubation, allowing for mistake-based learning without risk to patients.

Data Quality and Conversational AI

• 📈 Data harmonization and ensuring high-quality, complete, and aligned data sets are critical for effective conversational AI interfaces.
• 🛠️ Adverity's platform includes data quality components to monitor issues like naming conventions and data type alignment, harmonizing data (e.g., in UTC) and performing necessary transformations.
• 🎯 Designing conversational products involves defining the scope of supported features and selecting appropriate Large Language Models (LLMs) for different tasks (e.g., query compilation, response generation).
• 💬 Iterative development, benchmarking, and continuous testing of LLM responses are essential for improving the quality and reliability of conversational AI.