Uncover This Tech Term: Agentic Artificial Intelligence in Radiology

Abstract

Recent advances in artificial intelligence (AI) have been driven by the development of foundational models (FMs) [1].Large language models (LLMs) are FMs pretrained on a large corpus of text data followed by fine-tuning specifically for instruct-tuned models, with remarkable capabilities to 'understand' and generate natural language conversations [2].Extending beyond text alone, large vision-language models (LVLMs) integrate visual understanding with linguistic skills, enabling the simultaneous processing of both textual and imaging data [3].Further leveraging the advanced reasoning capabilities of recent-generation LLMs, newly developed technologies have led to the emergence of agentic AI systems, which differ from their predecessors by possessing 'agency': the ability to 'observe' their environment, 'plan,' and 'act' autonomously [4].Relevant terms are summarized in Table 1.AI agents leverage FMs as their cognitive core-or "brain"-to communicate, reason, and make decisions.A crucial distinguishing feature of agentic AI is its capability to engage in a plan-action-observation cycle.

Ähnliche Arbeiten