←  Back to overview
Article

The Dangers of ChatGPT and Other Generative AI Technologies

Generative AI is all the rage, but this surge in mainstream appeal is not without its own set of issues. Our content lead weighs in.

Foreword

During my time at Bothrs so far, I've written hundreds of posts, dozens of articles, and a bunch of white papers. And to me, seeing what ChatGPT can conjure up in the blink of an eye is equal parts incredible and foreboding. I actually tried to let the AI do my job, and while the results weren't exactly mind-blowing, I could very well see them being passed off as "good enough" by many. Exciting stuff but also a little concerning, for more reasons than one.

The virality of AI technology

If you haven't heard of ChatGPT by now, well, where have you been? It's the latest AI experiment by OpenAI, which accrued well over a million users in just five days. For perspective, it took Facebook 10 months to reach that number. If you don't know, ChatGPT is an AI chatbot. Ask a question, and the AI answers. Tools like GPT-3 and DALL-E 2, the breakthrough image generator that came out earlier this year, were also created by OpenAI.

GPT stands for “Generative Pre-trained Transformer”, by the way, and people have been using it to create recipes, draft legal agreements, write emails, blog posts, skits… I mean, people were even writing code with the help of the chatbot. Absolutely bonkers. If you want to get a glimpse of what people have been up to, this article by the New York Times sums it up it pretty well.

ChatGPT is already being heralded as the new Google because of its seemingly boundless abilities, but unleashing this technology onto the masses is not without caveats from where I'm sitting. Let's explore.

Is this town big enough for the both of us?

Today's generative AI essentially mimics what digital creatives do, ranging from digital art and designs, to copywriting and content creation. And they've gotten pretty good at it. So if I were to summarize the potential dangers of AI for digital creatives with regards to their livelihood, I would compare it to Gresham's Law. It's an economic principle that states “bad money drives out good money”. Meaning, if you got two commodities that are both accepted as having similar value, the undervalued commodity will gradually phase out.

If original content is costly but higher in value, and free but mediocre content can be manufactured on a whim, the latter will likely thrive. I think the moment we start collectively undervaluing human creativity, AI-generated content may have the potential to drown out content created by humans as time moves on and technology progresses.

Take ChatGPT as an example. From my perspective as a marketer, just the sheer volume of content you'd be able to publish with the help of artificial intelligence is probably enough for many businesses to re-evaluate their content strategy. Everyone knows it's only getting more difficult to grab people's attention on the internet. And if you can't do that by standing out, your other option is to just bombard people. Content up the wazoo; extreme quantity over quality. The latter is becoming increasingly viable now thanks to AI, since it's:

  • Pretty much free or extremely low-cost
  • Requires virtually no skill to operate
  • Instantaneous compared to human efforts

I mentioned this in the intro; the output of generative AI right now is far from perfect though. At least, compared to what a human field expert could do. Both in terms of usability and accuracy. How are you going to deal with client feedback on design assets generated by an AI? And, just because you're using ChatGPT or similar, that doesn't automatically mean your information is factual. More on that later.

Hey, remember that AI-generated rapper from a while back, the one that somehow landed a record deal? That project got shelved because the AI became racist. Yeah. We've still got a long way to go until AI gets truly competitive. Though it's never too early to start worrying, am I right my fellow overthinkers?

Limitations and drawbacks of generative AI

Now you're probably wondering why I'm being such a Debbie Downer about this. Like, how is this a bad thing? Writing is hard, creating art even more so. Aren't we just making things more accessible? Well, yes, but the problems arise when you start looking at how generative AI works:

  • Data used to train AI models is often obtained illegally: Digital artists in particular are vehemently opposed to AI. For instance, to generate an image, the engine scrapes artwork all over the web and does so without any kind of permission. Anyone with their IP out in the open can have their shit stolen by an algorithm is basically what I'm saying. And it's not like you can file a DMCA claim against an AI. Additionally, “AI artists” are flooding showcase sites like Artstation and DeviantArt with AI-generated content, making it even harder for actual artists to get noticed.
  • Everything AI creates is inherently derivative: While yes, art does not exist in a vacuum, that does not put AI on the same level as human creativity. Art in itself is a human, personal expression, whereas AI will just create an amalgamation of expressions it found on the internet. This is very noticeable in digital art; most AI art just looks like a mash-up of art styles. And with ChatGPT, you're usually getting generic, academic language, with very little pizazz or depth. Artificial intelligence can't write what hasn't been written dozens of times before.
  • Answers given by an AI are not necessarily correct: To illustrate, the use of ChatGPT is currently banned on Stack Overflow, the popular Q&A website for programmers. The problem isn't just that ChatGPT can get it wrong, but rather the answers it gives you look believable. Experts and moderators now have to sift through floods of AI-generated content to check their accuracy. Combating misinformation in this day and age is already an insurmountable task; we really don't need more tools to spread BS even faster.
A guy on Twitter proudly proclaimed he made a children's book using nothing but generative AI over the course of a single weekend, and published it on Amazon for people to buy. Spoiler alert, people weren't having it.

Closing thoughts

Don't get me wrong, it's unironically amazing what ChatGPT or generative AI in general is capable of nowadays. For research or inspiration purposes, it's a fantastic tool to have in your arsenal. And as with every major technological disruption, it feels threatening to the old guard. But when money becomes a factor, and it always does, a lot of people won't take issue with casting aside their integrity for the sake of cheap clicks. I, for one, am not excited for the inevitable flood of AI-generated SEO content that's coming, further deteriorating the quality of search engine results for everyone.

I guess people flock to AI because of the vast amounts of output it can generate based on a simple text prompt. And yeah, it can really make your life easier at times. Also because it's just fun to play with. But it's not entirely harmless, especially to the people that created the data these models use to train themselves. The copyright issues here alone will make anyone's head spin. Artists and creatives can either get plagiarized, scrutinized, or both. The public perception of what it means to be an artist is already changing.

My only wish, is that people use AI for enrichment and personal use. Not as a shortcut or a means to manufacture exposure or monetary gain through blind appropriation of existing IP. The technology itself is incredible, but unfortunately lends itself a little too well for less kosher purposes in my opinion, especially in a system that puts individual gains over the welfare of a community.

Credit to @teacupbun on Twitter.

Join Bothrs

Bundle forces and let’s create impactful experiences together!
See positions