PicoBlog

Update on March 19, 2024: Please scroll to the bottom of this post for an update on the vetting process.

Writer’s note: There is another AI editor job posting called Outlier, which has an uptick in posts. However, before they’ll allow you to complete a test, they ask for your driver’s license and other confidential information. I don’t recommend releasing your license or any other documentation that can easily make you a victim of identity theft. Meet the staff. Talk to a human being. Background checks should come AFTER you get an offer.

If you’re a freelancer who is looking for new writing assignments, it’s all but impossible to miss the uptick in artificial intelligence jobs on career (and work-from-home) sites. I’d already learned a considerable amount related to artificial intelligence writing after a law firm client sent me a massive work log of Frequently Asked Questions (FAQs) to review.

I couldn’t put my finger on why the content read differently, but something was off. I knew his writing style, and these FAQs felt mechanical. The words on the page were factually correct (for the most part), but there was an impassive tone in the writing. I finally asked him, “Did you write this?” When he told me he was using an AI program, it all made sense. Now I could approach the editing in a different manner.

Imagine having standing water in your bathroom sink and a clog that Drano can’t fix. Now imagine the way a human plumber would talk to you about getting your sink back to working order versus reading an instruction manual for a drain plumber. The first option can answer questions in real time based off of lived experience (and shame you for using Drano). The second option couldn’t care less about whether you can wash your hands in this porcelain fixture ever again.

And that’s what it’s like to edit AI writing. AI editors are looking at punctuation, grammar and pacing, but they’re spending even more time on tone.

After months of editing AI material for the law firm, this DataAnnotation 45-minute test was giving me deja vu — with one exception. Instead of reading diplomatic legal material and financial advice, I was reading about politics and vetting whether certain forum comments would be considered free speech versus an immediate ban from a forum. You can’t just tell the test, “This person is a jerk, but this other person is making valid points.” You have to explain feelings to this software.

Recommended Read: “BlackTechLogy: Let's talk tech ~ An archive of all tech-related posts on ‘I Do See Color’"

By the time I’d finished the unpaid 45-minute test (at $20 per hour if hired), I was honestly too tired to take the second HTML/CSS coding test ($40 per hour if hired).

I am forever grateful for Bitwise Industries’ free courses in “Websites for Beginners,” “Mobile-Friendly Websites,” “JavaScript for Beginners” and “React.” I’m even more grateful to have taken one of those classes with a woman of color (Latina) and an assistant who was also Latina. Nowhere else in my 19-year history of working in digital editing have I seen tech training with one minority coder nor a woman, never mind two of both. Regardless of the board drama I just learned about, Bitwise will ALWAYS be good in my book.

I admit that I almost dozed off during the test. Unlike the paid legal AI editing work, this application is on a timer. Taking breaks to walk around and give your brain a break is not an option. I argue that it should be for material this dense.

While I understand why writing tests are often necessary for long-term jobs, I think paid tests will always get better results. It means the hiring manager has carefully chosen job candidates instead of trying to vet people through a mouse maze of personality tests and multiple-choice questions (that too often aren’t updated with Associated Press’s ongoing style changes). And by the time you finish those unpaid job tests, even less-competitive job seekers feel like they wasted their time.

So would I say taking the DataAnnotation test I took in January was a waste of time? What happened after the test was completed?

ncG1vNJzZmihlKTAprHCqKOoql6owqO%2F05qapGaTpLpwvI6dmK2ZXZa7r7vTmquip55it7Cu