First Principles AI

Recently, someone sent me a survey to be filled in about the use of ChatGPT among doctoral students. Looking at the survey it was clear that the researcher had already formed an opinion - ChatGPT was cheating and human work was not! And I’m not surprised because, since the advent of ChatGPT, plagiarism has become a prime concern of universities. On one hand, educators want students to demonstrate originality. On the other, students have to sort through so much information that after a while they subconsciously lose track of whether the ideas are truly their own and what ideas they've taken from others. Many rationalize that it was ‘their’ idea but written by an AI.

This has also led to an ever-blowing spiral of language models vs detectors. Students are subscribing to more AI tools that ‘humanize’ output and educators are 'subscribing to ‘AI plagiarism' detectors’ in response.

However, the fundamental question still remains - whether AI writing should be considered plagiarism at all. And why did the researcher (who sent me the survey) fail to look at the other side of the story: what if AI Content = plagiarism is just a social construction; a norm of a bygone age where only manual writing was considered original?

In fact, pry a little deep and you’ll see that what becomes ‘plagiarism’ in academic settings becomes a bestseller book in another!

 The NYT Bestseller ‘Steal Like an Artist’

The question is how can we examine and develop multiple perspectives on issues such as these when developing research questions?

This is where the concept of ‘First Principles thinking’ comes in.

First Principles thinking is about dissecting complex problems into their simplest, fundamental concepts through logical reasoning to discover optimal solutions from new directions. It begins with pinpointing the main problem statement, listing all hurdles, critically questioning these obstacles, extracting the fundamental truths, and then building on these principles to unveil new solution avenues and tackle the problem efficiently.

Historically, the concept of first principles is not new. It has been a cornerstone in various fields, aiding in breaking down complex ideas into their most basic form. First Principles Thinking, for instance, has been a powerful tool for scientists and inventors (most famously Elon Musk), allowing them to strip away assumptions and build understanding from the basics.

Now, while the task of aligning AI using first principles predominantly rests with engineers, the ethos of First Principles Thinking isn’t confined to them. Researchers, too, can employ this method to enhance their inquiries. It helps in dissecting complex issues, questioning prevailing assumptions, and fostering a deeper understanding of the subject matter.

Here’s how you can use First principles thinking using ChatGPT:

Prompt: "I want to dissect the fairness of treating AI-generated content as plagiarism using First Principles Thinking. Start by identifying the fundamental truths or assumptions, challenge these assumptions, and then reconstruct the understanding based on the core principles. Provide a detailed analysis."

Here’s an example of the plagiarism issue:

Fostering a culture of inquiry and deeper understanding is a habit and AI can be a great help in it. The synergy between thinking frameworks and AI can provide a fresh lens through which we view, interact with, and shape our own minds and develop an understanding grounded in core truths and principled reasoning.

Till next time!