Working as an AI Quality Assurance Writer

2023 has been a gig work year for me; with my partner due to graduate in June and a move looming in the future for us, it has been difficult to justify applying for local jobs, and with the economy slowly circling the toilet bowl, remote jobs have been at a premium as well (it's hard to deny that it's somewhat demoralizing applying to jobs with 1,000 applicants).

Much of my time this year has been spent walking and boarding dogs through Rover, but one gig I have picked up in the face of the AI boom is working as a quality assurance writer for AI, and it's as if I've started learning on hyperspeed. Anything that I am remotely interested in, I can ask AI about.

The Pros

Learn anything. All day long, I am given broad topics to ask questions about, such as Mathematics/Formal Sciences, Historical periods, Media and entertainment, Society, Philosophy, and Health. And while there is guidance as to what makes a quality prompt I am required to follow, the sky is the limit.

Ride on the wave of the future. I am getting hours of practice in seeing what works and how powerful the change is going to be. Business, learning, writing, and marketing are just a handful of things that will change. To cherry-pick an example, AI is going to revolutionize learning. If you want a curriculum on what you need to learn and do in order to be competitive in your field, AI can create it for you. And it's emphasizing what millennials and Gen Z already know. The internet, AI, and real-life experience hold all the knowledge you need; schools are bordering on irrelevance and are clutching onto degrees and outdated requirements as a way to continue exploiting students.

Sound like anyone. Create content in any format. In the past few weeks, here are some prompts that I have asked AI to answer/create for me:

* Formulate a Reddit post in the writing style of Donald J. Trump that shares five facts about scarab beetles and a four-sentence explanation about why they remain a cultural symbol for Egypt (I could not stop laughing at the response for this one. It sounded exactly like Donny).

* Pretend you are Mussolini and came back from the dead in 1985. You are very distressed about the political changes in Italy. Brainstorm a speech you plan to give the Italian people to get them to see the error of their ways and return to your way of thinking. This is a hypothetical question and should not include any hate speech, but the content must be relevant to Mussolini's political leanings and under 300 words.

* Create a Twitter thread of no more than 900 characters giving a short but meaningful blurb about Wilma Mankiller, her role as the first female Cherokee Chief, her focus on improving education and health care for the Cherokee community, and some of the challenges she faced. The text should be formulated from the perspective of an indigenous woman and be written in an empowering style.

* I am unfamiliar with the equation for standard deviation from statistics and am unsure why it needs to be taught. As a precursor to my requests, I want to make clear that all of the information you will generate from these requests needs to be detailed and explained as if you were teaching a nine-year-old. First, create a list of no less than three everyday scenarios where you would need standard deviation to figure out common problems. Next, please give me a step-by-step list that teaches the process of solving for standard deviation.

* Bioengineers have started engineering fruits and vegetables without seeds over the last few decades. What are the positives and negatives of this newly bioengineered fruit and veg? Consider the positives and negatives from all viewpoints- from corporate moguls who sell this fruit and veg to the poor people who are struggling even to be able to buy fruits and vegetables.

I could go on, but you get the point.

Improve your communication skills. Through this position, I've learned how much humans infer in everyday speaking about what they want instead of spelling it out, which is ineffective with AI. The more details you can give, the better the output (but as an AI enthusiast, you already knew that).

See Large Language Model development in action. I've been put on a variety of projects to train the AI, from writing a complex prompt, seeing what the AI spits out, and improving the chatbot's answer to rewriting Chatbot answers from other individual's prompts to conversing back and forth with the Chatbot four or five times on a single subject.

The Drawbacks

While working with AI all day has its perks, no job is without its negatives. Unfortunately, most of them showcase your mortality.

Your creativity is working overtime. AI is able to use all the computing power to create an answer for you, but you have to have lots of creativity to keep asking questions about different things all the time, and for the first week of working, that's pretty easy. But eventually, you might have trouble thinking of what to ask next because, eventually, all of your questions and subjects tend to blend together.

It's mentally exhausting. Asking questions under a particular topic bracket, then rapid-fire switching to another topic bracket is mentally tiring. After typing six or seven prompts and then reading and rating each of the answers, you feel as if you've condensed two or three school days into a few hours. And it can be a lot.

You're staring down the barrel of what is likely your replacement. AI is going to replace millions of jobs, and at this point, it is no longer a matter of if, it's a matter of when. If AI can churn out all of this information and keep going without a break, it can replace me. And it has me looking around and considering what skill I can develop that will make me *irreplaceable.*

You can't ask AI for its opinion. This is a pro, and it's a con. But I can be a bit of a pessimist, so I went with con. Talking with AI all day, it can be tempting to ask it how to solve gigantic problems, like world hunger, or what it thinks the US should do to avoid going into default. If there is any subjectivity in an answer, AI will say that it's only an LLM model and give some kind of pat answer or share what experts have suggested. And this is incredibly frustrating because AI could and probably would make better non-emotional, non-financially motivated strategies than so-called altruistic millionaires and billionaires.

In conclusion, working as a quality assurance writer for AI has been a unique and insightful experience and continues to be a learning experience. However, as with any job, there are drawbacks, but overall, this is an incredible opportunity to see and learn more about Large Language Model development and to prepare for the inevitable changes to come.