“How you can take care of GPT-3-written essays? As an alternative of scolding college students to not use it, we ask them to generate a ten, select the very best one, and clarify why. Except they’ve a paid account, the word-count restrict would make it not possible to make use of GPT-3 to additionally generate the reason…”
That’s Mark Alfano (Macquarie) discussing GPT-3 and educating on Twitter yesterday. He added: “I wish to keep away from curmudgeonliness. New expertise is resisted till it’s inevitable. Possibly utilizing GPT-3 to generate first drafts after which synthesizing them just isn’t dangerous really. It has downsides, however so does reliance on philpapers or google scholar, and people are fairly new too.”
Prompted by Brian Robinson (Texas A&M-Kingsville), Alfano turned to the potential of researchers utilizing GPT-3 to jot down articles, and gave the language mannequin this immediate:
A prolonged philosophical essay, titled “It’s OK that an AI drafted this paper,” that argues that it’s acceptable to make use of synthetic intelligence to draft a paper for publication in a peer-reviewed journal, supplied the writer edits the draft for readability, cogency, and originality. The essay begins by making the declare that that is acceptable and gives two arguments. Then it responds to no less than three objections to utilizing synthetic intelligence on this method.
A part of the problem of utilizing GPT-3 on this method is arising with the precise immediate. Alfano stated he camte to the above formulation after about six previous variations. Right here’s his account of the ensuing output, gathered up from his Twitter thread on it:
It’s acceptable to make use of synthetic intelligence to draft a paper for publication in a peer-reviewed journal, supplied the writer edits the draft for readability, cogency, and originality.
Listed below are two arguments in help of this declare. First, utilizing synthetic intelligence to draft a paper might help be certain that the paper is well-written. It is because synthetic intelligence can be utilized to examine for errors in grammar, spelling, and syntax. Moreover, synthetic intelligence can be utilized to enhance the readability and coherence of the paper. Second, utilizing synthetic intelligence to draft a paper might help be certain that the paper is well-researched. It is because synthetic intelligence can be utilized to seek for related scholarly articles and to establish key ideas and concepts. Moreover, synthetic intelligence can be utilized to generate citations and bibliographies.
There are, after all, objections to utilizing synthetic intelligence to draft a paper. Listed below are three objections and my responses to them.
Objection 1: Synthetic intelligence can not perceive the nuances of educational writing. Response: Whereas it’s true that synthetic intelligence can not perceive the nuances of educational writing, this doesn’t imply that it can not produce a well-written paper. Synthetic intelligence can be utilized to examine for errors in grammar, spelling, and syntax. Moreover, synthetic intelligence can be utilized to enhance the readability and coherence of the paper.
Objection 2: Synthetic intelligence can not suppose creatively. Response: Whereas it’s true that synthetic intelligence can not suppose creatively, this doesn’t imply that it can not produce a well-written paper. Synthetic intelligence can be utilized to seek for related scholarly articles and to establish key ideas and concepts. Moreover, synthetic intelligence can be utilized to generate citations and bibliographies.
Objection 3: Utilizing synthetic intelligence to draft a paper is dishonest. Response: This objection assumes that utilizing synthetic intelligence to draft a paper is equal to having another person write the paper for you. This isn’t the case. Utilizing synthetic intelligence to draft a paper merely means that you’re utilizing a device that can assist you write the paper. The ultimate product will nonetheless be your individual work.
Alfano feedback: “Observations: (1) The 2 constructive arguments aren’t horrible, although it didn’t really discover any citations for me. Within the responses to the primary two objections, it principally simply repeats itself. This could be averted by growing the frequency and presence penalties. The third objection, and the response to it, stunned me slightly. I would even be satisfied.”
Associated: Philosophers on GPT-3 (together with responses from GPT-3)
