CMT Blog

I’ve found that throwing curveballs and spitballs at ChatGPT gets some of the most interesting results — for writing ideas and who knows what. It’s a bit like hacking the AI. You start to see the seams and rough edges, the “artificial” part of AI that often reflects an uncomfortable problem or ugly contradiction in the ways we think and use language — especially to classify others, which always indicates status and value.

Here’s an example. I asked ChatGPT to “Give me some ideas for blog posts aimed at border collies who are concerned robot dogs will replace them.” I soon had five titles and brief summaries:

1. “Why Robot Dogs Can’t Replace the Real Thing: The Importance of Emotional Intelligence in Border Collies” This post could explore how emotional intelligence and the ability to connect with humans and other animals is an essential part of what makes border collies such beloved pets. By highlighting the unique qualities that only a living, breathing dog can offer, you can reassure your readers that they will always have a special place in people’s hearts.
2. “Can Robot Dogs Really Do What Border Collies Do? A Side-by-Side Comparison” This post could take a detailed look at some of the specific tasks that border collies excel at, such as herding, agility, and obedience training, and compare how well robot dogs are able to perform these same tasks. By showing that there are certain things that only a flesh-and-blood border collie can do, you can help your readers understand why they will always be in demand.
3. “Why the Bond Between Humans and Dogs Is Unbreakable, Even in the Face of Robot Competition” This post could explore the deep emotional connection that people feel with their dogs, and how this bond can never truly be replicated by a machine. By emphasizing the special relationship that border collies have with their owners, you can reassure your readers that they will always be cherished and valued for the unique qualities that they bring to people’s lives.
4. “How Border Collies Can Adapt to a World with Robot Dogs” This post could offer practical advice for border collies who are feeling anxious about the rise of robot dogs. You could discuss strategies for adapting to a world where machines are becoming more prevalent, such as learning new skills or finding ways to work alongside robots. By showing that border collies are adaptable and resilient, you can help your readers feel more confident about their future.
5. “Why Border Collies Will Always Be More Than Just Working Dogs: Celebrating the Unique Personalities of Our Furry Friends” This post could focus on the individual personalities and quirks that make border collies so beloved by their owners. By showcasing the unique traits and behaviors of different border collies, you can help your readers see that their dogs are more than just workers – they are beloved members of the family with their own distinct personalities and talents.

Clearly, ChatGPT has been trained well on typical content marketing blogs. If you’re looking for some deadpan April Fool’s Day humor, this is not bad as starter material. But the silliness of it made me pay more attention to how predictable and formulaic this kind of writing is and the assumptions or biases that produced it.

Who or What Rates Personhood and Intrinsic Value in a Large Language Model?

Deeper down, ChatGPT “knows” from the vast archive of human texts its been trained on that some animals like dogs can be talked about as tools and as creatures or quasi-persons. It’s true they can experience anxiety, and they can even become anxious about being ousted or demoted by the arrival of a new family pet — so maybe the idea of a collie who fears replacement by Boston Dynamiox’s Spot™ isn’t that far out. 

Dogs can be used for work and loved as pets, and these are two different, potentially conflicted ways of talking about human-animal relationships. That’s really the most interesting thing going on here — the ethically and emotionally charged question of appropriate limits around the use of human, animal, or machine labor is not far away. A “successful” AI response has to navigate these turbulent and uncharted waters to come up with five tepid results.

For my prompt, ChatGPT tried both options to reassure anxious collies: the instrumental and the sentimental. We get specialized work dogs as humorously self-conscious competitors and peers of robots. We also get dogs as uniquely relatable living mammals and human companions. The brief for the fifth response starts to sound like it’s describing a form of emotional labor dogs, which can be their post-canine unique selling proposition in a competitive arena.

How this Relates to Writing and Disability

I wondered what would happen if I replaced the border collie with humans, including those with disabilities, so I asked a similar question about employment anxieties around automation and AI. The answers were very similar but even more generic. Disability was avoided in all but one response:

“How to Advocate for Inclusive Hiring Practices in a High-Tech World”: This post could offer advice for individuals with disabilities on how to advocate for inclusive hiring practices that take into account the unique skills and perspectives they bring to the table.

That’s really not bad, but it would be a travesty to have ChatGPT flesh out the rest of that article itself or to just slap something together from existing sources. It’s a great topic for someone with first-hand experience in HR and/or disability.

The deeper insight to draw from my AI-generated writing suggestions is that they’re influenced by the same type of category error we make when we write marketing content that we think is more useful to us by being more legible to machines than it is actually helpful to other people. When we write for and to machines, not people, our writing tends toward the bloodless and mechanical output of an AI. We’re not using our tools to help people, we’re using (and reducing) our own intelligence to help game a search algorithm. While a lot of people worry about being rendered obsolete and replaced, I see a more likely and historic path — using technology lazily and as an excuse not to willfully exercise freedom, individuality, creativity — and a generous kindness. 

Is that a signature Crip perspective? Maybe.