Review of: "[Commentary] Are Large Language Models obsolete? And are multimodal solutions inappropriate?"
Shahid Ikramullah Butt · 2024
Potential competing interests: No potential competing interests to declare.Conclusion recommends, "principles of lateralization, sensory-motor integration, gestural syntax, and embodied cognition, these agents can achieve more human-like interactions, better adaptability, and more effective learning capabilities."These improvements, if implemented, can really support a more realistic artificial intelligence environment, which can be close to the human response and also save much time for humans.Specific examples of how principles like sensory-motor integration could be applied in AI design: for example, if a real-life problem where SWOT analysis is required arises, this technique can give some guidance to make a calculated decision.The proposed dual-network architecture, which incorporates two distinct neural networks (one for LLM and one for gestural language) with a metaphorical translation apparatus between them, might address current limitations of LLMs in several significant ways:1. Enhanced Multimodal Processing Current Limitation: LLMs primarily process text and struggle with integrating other modalities like gestures, images, or sounds.Proposed Solution: By having separate networks specialized for different modalities (LLM for text and another for gestures), the system can more effectively process and integrate information from diverse sources.This multimodal approach can lead to richer, more nuanced interactions and understanding. Improved Contextual UnderstandingCurrent Limitation: LLMs often lack the ability to understand context derived from non-verbal cues, which is crucial in human communication.Proposed Solution: The gestural network can process and interpret body language, facial expressions, and other non-verbal cues.When combined with textual information through the translation apparatus, this can provide a more comprehensive contextual understanding, leading to more accurate and human-like responses. Creative Problem Solving and Error HandlingCurrent Limitation: LLMs can generate plausible-sounding but incorrect or nonsensical answers, and they have limited creativity in generating novel ideas.