• ChatGPT: AI versus Me in Writing an Opinion Column

    February 3, 2023 // 3 Comments »

    ChatGPT is an online tool created by the Open AI people that generates text. You prompt the software and out pours decent prose. Someday it is going to replace all of us, starting with people like me, we are told. So how good is it? Let’s see if the software can recreate a recent column from TAC. In the original hand-written column I presented four global memes with commentary. Let’s include the memes below in italics and see what CHatGPT has to say about each.
    The internet promotes democracy globally. I called the idea a grift and offered Twitter’s demise as a platform for free speech as an example. Here’s what our future overlords at ChatGPT had to say: “The internet has the potential to promote democracy and increase access to information around the world. By allowing people to connect and share ideas with others, the internet can facilitate the exchange of diverse viewpoints and facilitate the free flow of information. In this way, the internet can help to create more informed and engaged citizens, who are better equipped to participate in the democratic process. However, it is important to note that the internet also has its limits and can be used to spread misinformation and propaganda, which can undermine democracy. It is also important for individuals to use the internet responsibly and critically evaluate the information they come across online.”
    Points = 0 for sounding like a weak Tom Friedman column, and – 567 points for the fear-mongering bit about undermining democracy like a Democratic Party talking point. But at the same time it sure sounded pretty didn’t it? But I’m not worried about my job at this point.
    Countries with a McDonald’s don’t make war on one another. It was in fact that ever-so-journalistic New York Times columnist Tom Friedman who coined the phrase. He wrote the benefits of economic integration reduce the policy choices open to governments, making war—which disrupts that integration—so unattractive as to be practically unthinkable, part of all that end of history stuff that was once the vogue for the hive mind. The concept was built around everyone wanting to be more like us. It failed miserably. ChatGPT said “This statement, known as the ‘McDonald’s peace theory,’ suggests that countries that have McDonald’s restaurants are less likely to go to war with each other. However, this theory is not supported by empirical evidence. While it is true that McDonald’s and other international fast food chains are often seen as symbols of globalization and may be viewed as promoting international cooperation, the presence of these restaurants does not guarantee peace between nations. There are many other complex social, economic, and political factors that contribute to the likelihood of a country going to war. Additionally, the presence of McDonald’s restaurants in a country does not necessarily reflect the overall level of economic development or political stability of that country.”
    Here’s 567 points for laying out the theory more clearly than I did, and noting the complexity Friedman and I try to hide behind in our self-mocking/mocking. Clear advantage for ChatGPT.
    You have to be at war to be at war. I was going for the sarcasm in writing how U.S. weapons animate the conflicts with Ukraine, Yemen, and Syria, and U.S. assistance runs like code in the background of the Israeli-Iran power struggle across the Gulf. Sure sounds like war. ChatGPT misses the boat entirely by getting hung up in semantics and missing the specific examples, especially Ukraine: “This statement is not entirely accurate. War is generally understood to be a conflict between two or more countries or groups, often involving the use of military force. However, a country or group does not necessarily have to be in a state of formal war in order to be considered to be at war. There are many other types of conflicts and hostilities that can exist between countries or groups that do not meet the definition of war. For example, there may be ongoing disputes or tensions between countries that do not rise to the level of armed conflict, but which could still be considered to be ‘at war’ in a metaphorical sense. Similarly, countries or groups may engage in acts of sabotage, espionage, or other covert operations that are intended to disrupt or harm the interests of another country or group, even if these actions do not meet the formal legal definition of war.
    No points to either side for being predictable, albeit in opposite directions. Advantage me only because the Matrix bit expands my argument where ChatGPT is still checking Webster’s.
    U.S. leadership is near dead. That sarcasm thing again. I use examples from Covid, China, climate change, and NATO/Ukraine to make the point few nations are looking to the U.S. for global leadership on these global issues. ChatGPT, which appears to be programmed to avoid hard opinions, an unfair disadvantage in this contest, comes off weak: “This is a subjective statement that is open to interpretation. Different people may have different opinions about the strength and effectiveness of U.S. leadership in various contexts. It is important to note that the United States is a complex and diverse country, and its leadership and influence are expressed in many different ways. The United States is a major global economic and military power, and it plays a significant role in international affairs. However, it also faces challenges and criticisms from within and outside its borders, and its actions and policies are not always supported or welcomed by other countries and groups.”
    No points awarded once again because it appears ChatGPT is programmed to present a wishy-washy both sides argument. And who can argue with that? Advantage again to me because I am the one judging this.
    So while my original column eked out a one-on-one win here, it is obvious ChatGPT is a powerful tool. Imagine if the assignment was some simple blog text, or a high school level term paper on the “Causes of the Civil War” (I just tried it — ChatGPT nailed it, listing slavery and states’ rights are the main causes, with a neat paragraph on other economic and cultural factors. B+ or better level paper, probably would pass at the undergrad level with some footnotes and additional examples.)
    As to the value of ChatGPT itself for opinion writing, it works best when instead of trying to trick it into goofing up you work within its limits to pry out its best. Just like knowing a grammar and spell checker are there to help the human editor, who still makes stylistic choices, cuts for length, and calls for additional information, ChatGPT is a helper for a writer at present, not a way to produce a finished product.
    Actually, it is best to allow the software to speak for itself: “AI in writing opinion columns can potentially save time and improve efficiency by automating certain tasks or providing assistance with writing and editing. However, it is important to note that AI systems are not able to fully replace human judgement and creativity, and it is still necessary for humans to review and approve the final product before it is published. AI can be a useful tool for writers and editors, but it is not a substitute for human expertise and insight.”

    Related Articles:

    Copyright © 2020. All rights reserved. The views expressed here are solely those of the author(s) in their private capacity.

    Posted in Other Ideas