Knowledge Tree
Relations (1)
related 0.50 — strongly supporting 5 facts
Large language models are a specific application of deep learning technology, as evidenced by their shared neural network architecture [1] and their common limitations regarding generalization and confabulation [2]. Furthermore, deep learning provides the foundational framework for feature learning in LLMs [3] and serves as the underlying methodology for applying these models to complex tasks like health record analysis [4].
Facts (5)
Sources
Understanding LLM Understanding skywritingspress.ca 3 facts
claimLarge language models (LLMs) possess the same basic limitations as other deep learning-based systems, specifically struggling to generalize accurately outside of their training distributions and exhibiting a propensity to confabulate.
procedureResearchers at McGill and MILA used deep learning to interpret clinician thinking by pre-training on hundreds of millions of general sentences and applying large language models to over 4,000 free-form health records to distinguish confirmed from suspected autism cases.
claimMisha Belkin from UCSD presented on dimensionality and feature learning in Deep Learning and Large Language Models at the 'Understanding LLM Understanding' summer school.
Not Minds, but Signs: Reframing LLMs through Semiotics - arXiv arxiv.org 1 fact
claimDespite modern Large Language Models (LLMs) not operating through symbolic logic, the metaphors of cognition have persisted and intensified with the rise of deep learning, with traces of the 'mind-as-machine' metaphor surviving in recent neural approaches.
Building Better Agentic Systems with Neuro-Symbolic AI cutter.com 1 fact
claimDeep learning neural network-based large language models, such as GPT-4, Claude, and Gemini, process unstructured data including text, images, video, and streaming sensor data to learn patterns, classify data, and make predictions.