In Part 1 of this series, I looked at some specific pushback to my posts that question the dominant Ai narratives out there. In Part 2 I took a step back and looked at pushback to what others have written about their concerns with Ai. In this post, it is time to turn my attention to the technology in question.

Some have said I am being too hard on Ai technology because there are good uses of it. Some have said that I have been too harsh on those that use Ai as part of their daily routine. One commenter on this blog decided to have Ai speak for itself by asking ChatGPT 4o to review a post he took issue with. The post in question is Ai: The Trend That Was Promised to Be Different Keeps Following the Path of All Other Fads. I think the author wanted to prove that while there are legitimate concerns about Ai out there, there is also some use for Ai… while also subtly telling me I was off-base in my criticisms. I will just quote the (de-identified) comment here and respond to each paragraph.

“I always enjoy reading your stuff! Call me more of a middle ground person – I think there’s validity to some of your observations and concerns (completely agreed on the “another tool in my toolkit” take), but I’ve observed far more benefit than I think you’re giving GenAi credit for. I’ve found considerable value and time savings in having GenAi draft up proposals, summarize articles so that I can decide if I want to spent the time reading them, recording meeting minutes and next actions steps in MS Teams (great for an guy like me with attention issues), making technical instructions more concise and reader-friendly, and more. But yes, there’s a lot of hype and fluff, too. I think much of CoPilot for M365 is garbage for a variety of reasons, and I’m really hoping Microsoft rethinks their design approach to this tool.”

I kept this part in tact just so you can see that this person uses Ai in ways that I have expressed concerns about by leaders in education (summarizing articles specifically).

“I’m fascinated by logical fallacies in arguments, and so I asked ChatGPT 4o to review your article and determine if any logical fallacies existed – see below. Not bad. Of course, knowing Ai is biased, you have to take these with a grain of salt. :>)”

So, obviously one could point out that blog posts are generally about expressing opinions and commentary, not logical arguments. Right of the bat here, using ChatGPT in this manner is just a major misunderstanding of how to use logical fallacies. My post in question is a rebuttal and alternate viewpoint (if that wasn’t obvious from the title), not a logical argument. However, if ChatGPT is so good at what it does, surely it would account for the differences between a logical argument and a blog post, right? This mistake does add to the concerns I have been expressing in this series. Let’s see how ChatGPT did with evaluating the post based on logical arguments.

The argument contains several logical fallacies, which detract from its overall validity. Here are some of the most prominent fallacies:

Hasty Generalization: The author quickly concludes that Ai is a fad, based on limited personal experiences and anecdotes from social media and friends. The argument relies on these selective instances rather than considering a broader spectrum of evidence. This fallacy is evident when the author cites that “most give it up” because it’s quicker to do things manually, without exploring the vast number of use cases where Ai has been beneficial.

Right away ChatGPT has already fallen apart. In general, I was responding to the idea that “Ai is different from past fads.” The assumption that Ai is a “fad” was already baked into the idea that I am responding to. Whether or not Ai is technically a fad is irrelevant to the point. But the part where Ai quotes me was where I am talking about my experience with those I follow and talk to regularly. I had pointed out it was anecdotal. Of course, that doesn’t mean I didn’t consider a broader spectrum of evidence. ChatGPT is just assuming an argument I wasn’t trying to make in the first place. But then again, part 2 of this series looked at how humans assume you don’t know what you are talking about if they disagree with your point, so I guess ChatGPT is just following it’s training data?

Cherry-Picking: The author selectively focuses on examples where Ai has failed or caused harm, ignoring successful and impactful uses of Ai in various fields. For instance, they dismiss Miguel Guhlin’s Ai breakthrough as an “outlier” and only cite failures, thus presenting an unbalanced view. The argument would be stronger if it acknowledged both successes and failures.

This one sounds good at first. But think about it: it’s not “ignoring” successes when you are specifically presenting a counter-argument. Also, I don’t just dismiss Miguel Guhlin’s point – I point out very specific ways that he didn’t make his point. I wasn’t making an argument – the point was to raise questions that Guhlin did not address. He pointed out his own successes, I was just presenting a counter balance. A human reader would probably get this.

Ad Hominem: There is an implicit attack on those who claim Ai is useful. For instance, the suggestion that people like Guhlin’s achievements are met with incredulity (“THAT took you weeks? Really?”) dismisses their credibility rather than engaging with the merits of their arguments.

Except I did engage with the merits of Guhlin’s and other’s arguments, so there were no ad hominem attacks. The second sentence in the quote above actually confuses the whole point of the part of my post it is referring to. I didn’t suggest such a thing, and even if I did – that statement technically did engage with the merit of their argument (questioning the massive time savings claimed).

Slippery Slope: The argument moves from discussing Ai-driven ordering systems to grandiose claims about Ai’s harm, such as environmental destruction and injury from self-driving cars, without a clear connection between these disparate issues. This makes it seem like any adoption of Ai will inevitably lead to these severe consequences, which is an oversimplification.

This one also sounds good at first, but it is pretty much gibberish to some degree if you think about it. The harms of Ai are documented, not “grandiose claims.” Ai is causing these harms, so there is a clear connection. It’s not oversimplification, it’s a clear causation.

Strawman Fallacy: The author misrepresents David Wiley’s point by equating “Generative Ai Education” to “Office Software Education,” thereby reducing a nuanced argument into something easily dismissible. The author also exaggerates the concept of “tool worship” and dismisses it without engaging with the broader point about generative Ai’s role in education.

This just shows ChatGPT has no idea what it is talking about sometimes. The concept of tool worship is not exaggerated – it does happen, and I was making a case for it happening here. This point is actually one of many ways to engage with the broader point of generative Ai’s role in education. But my exact quote was “Saying ‘Generative Ai Education’ makes about as much sense as saying ‘Office Software Education.’” So I wasn’t equating the two, just pointing out how either one doesn’t make sense in the given context. Also, it is unclear what the “it” in the last sentence of the quote is referring to – is it saying I dismiss the concept of tool worship itself? I didn’t do that. Did it mean to say I was using tool worship to dismiss generative Ai in general? Because I didn’t do that either.

Appeal to Emotion: The argument uses emotionally charged language to make Ai seem more harmful than it might be in practice. For instance, it emphasizes that Ai bots bully people into suicide and references the destruction of lakes, which are serious issues but presented in a way meant to evoke strong emotional reactions rather than providing a reasoned critique of Ai’s role.

Well, yes, blog posts do use appeals to emotion. This is where looking for logical fallacies on a blog post that was meant to provide counter-points falls apart. Of course I am trying to appeal to emotion on these serious issues. Saying that my words would cause a “strong” emotional reaction is quite the stretch, but we really need to stop acting like emotion is bad.

False Dilemma: The argument implies that Ai is either a complete failure or a tool for “cool kids,” ignoring the possibility that Ai could be useful in some contexts and problematic in others. This oversimplifies the discussion by presenting it as an all-or-nothing situation.

I didn’t imply that. But I think ChatGPT confused other people saying everyone must use Ai with my responses to this assertation.

Appeal to Popularity: The author suggests that because “many, many” people agree with Heather M. Ross’s critique, it must be valid, which is a fallacy. Popular opinion doesn’t necessarily equate to truth, and it overlooks contrary evidence from those who have found Ai useful.

These fallacies collectively undermine the strength of the argument, making it less persuasive and more ideologically driven rather than factually grounded.”

Actually, I went into 7 points that back up my agreement with Ross’s critique, so it wasn’t an appeal to popularity. Maybe the list confused ChatGPT?

I’m not really sure what this post will add to the Ai discourse. It was really just a fun exercise into picking apart a ChatGPT response presented as “not bad.” It was almost all bad. I was really hoping to find some parts of it that were good feedback, just to prove that I don’t dismiss all Ai out right. But… there just wasn’t much to the quality of the response the more I dug into it. Which has been my experience with most Ai output – and many others’ experiences as well (not an appeal to popularity – I have no idea how popular this experience is. Just that is does exist for a decent number of people). There are an astounding number of times when an Ai Cheerleader will give proof of a “good” Ai output that looks solid at first glance, but then falls apart the more you dig into it. It makes one wonder how deeply they examined said output in the first place?

Leave a Reply

Your email address will not be published. Required fields are marked *

Your post must include a link to this post.

How does this work?
Write a response on your own website and include a link to this post. Then submit your response’s URL here. Your response may appear here after moderation. To update or remove it, edit or delete your original post and submit the same URL again. Learn more about Webmentions.