Note: OpenAI leads the general AI space, but AI companies are developing deep research tools and experimenting with AI-powered academic searching in support of research. Perhaps you have faculty or students asking you to present these tools to classes.
How will AI tools affect our traditional bibliographic databases? Will we see GenAI being put into our search platforms? Can we stop it?
Artificial intelligence encompasses broader goals of simulating human intelligence, followed by machine learning, which focuses on learning patterns from data, and deep learning, which uses layered neural networks. These systems learn through supervised, unsupervised, and reinforcement models, each suited to different types of data and problem-solving. At the core of deep learning are neural networks, which enable advances in natural language processing which is a key area for libraries and research involving text analysis, summarization, and retrieval. Modern NLP is driven by large language models (LLMs), which rely on transformer architectures, tokens and introduce challenges like hallucination. Information retrieval has evolved through embeddings and semantic searching, improving discovery beyond keyword and controlled searching.
Issues of bias, fairness, and explainability are critical, especially in scholarly and clinical contexts. Evaluating AI requires metrics such as precision and recall to assess performance. Generative AI now enables creation of text, images, and audio, raising important ethical concerns. As a result, governance and policy addressing copyright, misinformation, and accountability are essential for responsible AI use in libraries and beyond.
Moving Beyond Chatbot Technologies
Three intertwined technical developments have moved artificial intelligence (AI) beyond traditional chatbot technologies:
1) Agentic AI — the capacity of an AI system to pursue goals across multiple steps, deciding for itself which actions to take, which tools to invoke, and how to recover from errors. Agentic AI aims to accomplish complex, often long-horizon objectives with minimal human intervention by planning tasks, using external tools, maintaining memory, and adapting to changing circumstances.
2) Multimodality and tool use — contemporary AI systems can process and generate multiple forms of information, including text, images, charts, audio, video, and PDFs. They can also execute code, browse the web, query databases, and interact directly with software applications and digital interfaces.
3) Persistent context and self-correction — the ability to retain and use large amounts of information throughout an extended task, maintain working notes and memory, evaluate intermediate results, and validate or revise outputs when errors are detected.
First-generation chatbots such as ChatGPT (2022) were able to draft an essay paragraph, answer questions, or assist with studying. These capabilities, while useful, were largely transactional and confined to a single interaction. By contrast, AI agents autonomously conduct literature searches, extract data from dozens of PDFs, write and execute analytical code, draft a manuscript, verify citations, and iteratively refine its work. Agentic-based AI systems represent a qualitative shift beyond conventional chatbots, creating new opportunities for research and productivity while simultaneously challenging existing academic and professional practices.
Note: ..."checks its own work" is not yet reliably true. Current agents can verify, critique, and revise their outputs, but they can also confidently validate incorrect conclusions. The more defensible distinction is not self-checking but autonomous execution of complex workflows..
Note: Please use your critical reading skills while reading entries. No warranties, implied or actual, are granted for any health or medical search or AI information obtained while using these pages. Check with your librarian for more contextual, accurate information.
