Creative Conundrums: Human Originality versus AI's Predictive Analysis

AI has taken over our world in a variety of ways, by harnessing the limitless potential of predictive analysis to unravel probabilities and forecast trends, allowing it to become a powerful agent in changing the world. This gives stiff competition to human creativity, on the other axis, which is driven by the impulses of intuition, feeling, and sentiment, leading to the looming question if AI currently has or in the future will have the capacity to replace human ability.

This requires a thorough investigation of both of these domains to foreground what each of them can and cannot do. While AI relies on algorithms, data functions, and neural networks to parse data, and offer predictions, humans are driven by experience, perception, imagination, and emotions, bestowing the latter with a sense of novelty and inventiveness that AI, no matter how robust its techniques of deep learning, will never be able to match. The reason behind this is the fact that human creativity is propelled by a sense of imagination, which is the horizon that blends disparate emotions and experiences to harmonize them and produce something truly new, that has no precedent before.

Thus, despite the fact that recent neural language models have demonstrably generated remarkable texts, that display structure of thought and fluency in language to a degree wherein it is not distinguishable from human text for the most part, there exist errors and pervasive problems in AI-generated text which allow us to think of the differences in writing style, semantics, structure, and syntax between human and AI-generated texts.

Studies have found several ways in which AI lacks behind human content. The most glaring example of this is the sheer inability of AI to create contextualized and specific content that is full of concrete arguments, data, and questions, even as it absolutely lacks the ability to think of novel research questions and ideate new arguments.

Further, in terms of text perplexity (PPL), which can broadly be understood as a measure of uncertainty in a text, text generated by large language models is definitely lower in terms of perplexity when compared with human-generated text. This is owing to the fact that the technique used by AI to generate content forces it to generate a text with high probability, whereas a word that follows another world is likely to follow it in most instances in the data used to train AI. On the other hand, human-written texts are diverse and heterogeneous, sometimes veering towards the absolutely erratic. Further, AI often generates fake citations and data, creating massive problems in its applicability in scientific research.

Therefore, analyzing the coherence, consistency, and argument strength of AI-generated text reveals the shadows that lurk over it, making it severely lacking in quality when compared with quality human writing.