Vents MagaZine Music and Entertainment Magazine
Related Articles
As technologies continue to advance at great speed, the creative process has undergone significant changes, particular due to multimodal models like GPT-4. As frequently reported, while the impacts are felt in numerous fields, it is in the academic domain that the most profound effects are being experienced. GPT-4 capabilities have advanced to the extent that the work it produces can be submitted by students targeting upper second class degrees as their own work. Consequently, it is natural that some will exploit this resource.
They choose to do so because of the convenience, ensuring that their education investment ultimately leads to a degree, despite the ethical concerns. However, students traditionally produced original work, where they either performed the reviewing/editing process themselves or sought assistance from a professional proofreading and editing service. Editing is critical, which means it must be questioned whether such AI models can produce a similar level of quality.
Arguably, regardless of their level of sophistication, AI models cannot match humans when it comes to editing. Firstly, language is highly nuanced, such that AI experiences challenges with determining the suitability of a word in the context of a specific sentence. For instance, while some words may seem to largely mean the same thing, their connotations may differ significantly, which AI is often incapable of discerning. Secondly, while certain words like ‘good’, ‘a lot’ and ‘very’ may appear to be acceptable, they are somewhat lacking stylistically and should be replaced with superior alternatives such as ‘outstanding’, ‘likeable’, ‘substantially’, and ‘extremely’, respectively. As AI disregards this issue, the generated work can be uninspiring, which particularly impacts areas in which creativity is key. Thirdly, there is lack of consistency regarding the application of certain punctuation rules, including parenthesis, dashes, semi-colons, and commas. Consequently, AI will be influenced by the flaws of its human programmers, leading to uncertainty in this regard.
The fourth issue relates to the construction of the work. For example, sentences can be formed in multiple different ways, such as by adopting the active voice, as in ‘I performed analysis on’, or the passive voice, as in “analysis was performed on…” The decision regarding which should be used can depend on the specific field, institution, or journal in which the work will be published, which AI does not consider. Furthermore, at the paragraph level, editors often make subjective judgements regarding whether a paragraph is coherent and flows in terms of whether something appears to be correct, logically connects with the following sentence, avoids becoming repetitive, and ends appropriately. When reviewing the work in its entirety, the editor must evaluate the integrity of the argument as a whole. This involves deciding whether the story being told is coherent and logical, and the constituent parts are clearly delineated with smooth transitions between sections and paragraphs, generally by using effective transition sentences. In this regard, editors use their aesthetic judgement in addition to seeking technical precision, which exceeds AI capabilities.
Fifthly, AI cannot determine whether certain expressions are appropriate as its awareness of social/moral conventions is lacking, particularly as they may differ culturally. For example, gender neutral language is frequently used, such that AI may be incapable of differentiating between the inappropriate generic use of ‘he’ when referring to an unspecified person and where it is actually used to refer to an individual. Furthermore, while editors could draw attention to the moral acceptability of certain arguments, such as denying the holocaust, AI is incapable of making these types of judgements, as its ability to understand abstruse concepts like values, beliefs and attitudes is limited.
Sixthly, AI lacks the ability to determine the validity of evidence for a given argument. To do so, one must have extensive knowledge of a specific field, and although AI does possess the capacity to achieve this, choosing which evidence is more appropriate is often dependent on the author’s subjective judgement, and a specialist editor is best equipped to evaluate this decision. Additionally, situations where names or dates lack consistency may not always be detected by AI, which a professional editor could resolve effectively.
Lastly, an editor will develop an overall opinion of the work, based on which they can make recommendations on areas of improvement, ways in which certain aspects could be expressed differently, and restructure it for maximum impact. They can also draw attention to any parts lacking in clarity. As AI is incapable of developing an understanding of the work’s inherent meaning, including the broader context in which it is situated, whether it is in terms of the course, field, or society as a whole, such assessments will be beyond its capabilities.
It is evident that technological advances in AI will enhance its capabilities exponentially. Nevertheless, editing is ultimately an inherently human process dependent on the abilities and traits of human editors, meaning that AI can never truly serve as a replacement.
