It is far from any thinking, it is computation based on given goals. By all means it is similar to computer program just with the NLP.
In context of LLM, is above statement valid?
The statement makes a valid point when considering certain aspects of Large Language Models (LLMs) like GPT-3 or similar models. LLMs operate using statistical patterns derived from large datasets, which is akin to computation based on given goals—these models aim to predict the next word in a sequence based on the input they receive. This process can be seen as similar to how a computer program operates, using a set of algorithms and rules to perform tasks, except that LLMs leverage natural language processing (NLP) capabilities.
However, the comparison also has its limitations. While LLMs mimic comprehension and can generate human-like text, they don't possess understanding or consciousness in the way humans do. Their "knowledge" is purely statistical, derived from the data they were trained on, and they lack any form of true reasoning or intentionality. They excel at pattern recognition and can produce outputs that seem insightful, but this is more a result of sophisticated statistical modeling than actual cognitive processing.
In summary, while the statement is valid in highlighting the computational nature and pattern-based operation of LLMs, it's essential to recognize that these models do not "think" in the human sense. They process language based on probabilistic algorithms, which is a key distinction from human cognition.