A Legit Article

Is AI coming for your job? What can you do?

By Marc Carson · Friday March 3, 2023

Above: Several California cows calmly go about their day, while an AI somewhere plots to take their jobs. Photo by Marc Carson

AI is really popular lately. As of this writing, there are already thousands of different AI tools for various purposes.

Still, there are a lot of big questions about how AI and the various AI tools will change our world. It got us thinking!

So, we tried out a number of different AI tools here.

With mixed results.

Here are the good results

First, let’s get the good points out of the way.

AI tools can really help in lots of areas. They can help you:

• Write detailed emails or messages, especially those that need to be professional in tone, or delicate, for example
• Provide some quick ideas to address a general problem
• Create illustrations or art, videos, and so on
• Write code…

Wait…writing code? That’s a big, complex part of what we do here!

We thought we’d better try that out.

We tried writing code with AI

We had to give this coding thing a try. Here are some of the results:

• AI coding tools really want to give you the actual code to use! There is usually a convenient “copy” button to copy & paste their code into your tools.
• AI-written code was a convenient helper in many cases. Though AI does act as if it can write very specific code, it was especially good at giving the gist of how the code should look.
• AI code helped us recognize the difference between knowing and doing. It sometimes told us, or showed us, exactly how to do things, but in some of those cases we still didn’t feel like doing them on our own! Funny how that works…
• AI code also made some huge and frustrating mistakes.

AI code didn’t magically make us better coders. But it can probably help you learn to code, if you’re interested.

AI tools remind us of clip art and fonts

Using AI tools reminded us a lot of clip art and fonts, which people often use in graphic design work.

Novice designers sometimes think that having access to tons of clip art, fonts, and other “creative assets” makes you a capable graphic designer.

In the past, there was even this fear: Would clip art, fonts, and design software put professional graphic designers out of business?

Sometimes you still see ads like this:

“Design it yourself! Buy our collection of 29,000 fonts for $29.95! Free clip art collection included!”

As it turns out, there’s more to design than just throwing fonts and clip art together. Owning this stuff did NOT make you a graphic designer capable of professional work.

(But it still feels good to have lots of fonts to use, right?)

Speaking of rookie work: The mistakes made by AI tools were also pretty interesting…

What kind of mistakes did AI tools make?

Here are some examples of those coding mistakes:

• AI code was often obviously incomplete! Like, it would sometimes just stop writing the code and act like it was finished. (You can tell it to finish the code though, and it will usually try)
• AI-written code often included some really simple mistakes that beginners would make. For example, sometimes it didn’t count up numbers correctly. Yes, you read that right!
• AI code took a LOT of shortcuts. In some cases, we were annoyed at having to clarify what we wanted by constantly changing the wording of our requests, so that it would do the work the right way.
• AI code tools even LIED to us.

Not even kidding about the lying!

One AI coding tool sent us on quite a wild-goose chase, convincing us we could write code in a way that didn’t even exist yet.

(We’re leaving out the technical details here…but if you know what an API is, one AI tool wrote extensive code for a fictional API, supposedly built by a team of software developers. Those developers were very amused to hear of its existence!)

“Here’s your code! Voila!”

Except no. No voila!

After having a total Oh-my-god moment, thinking we’d really been handed something great, a solution we’d wanted for a while, copy-paste-done, this was a painful realization.

We reached out to other coders to confirm that the code was total fantasy.

Eventually, we did have a bit of a laugh, thinking about how naive our expectations had been.

But imagine trying to use such fictional code as part of your job! Promising someone you can do it, and then nothing!

That would not exactly improve one’s working reputation.

(Maybe we did discover one easy way for AI to take your job away—just use AI for everything!)

Such frustrating experiences could possibly make you a better coder though…well…possibly, in the most annoying way possible.

Shortcuts are an AI specialty

One of our specialties here at Marc Carson LLC is building for the web in a professional way. We build websites and web applications that are effective, resilient, and reliable.

So naturally, we snorted out loud a bit at the AI when it messed up. No true professional is above a good snort!

Shortcuts are nothing new. But we don’t take shortcuts that others do, and that’s part of our reputation.

In many cases we refer to those shortcuts as “technical debts”. They will cause huge problems later, guaranteed.

(Did you think we wouldn’t attempt to make ourselves look good? Sorry, but AI tools seem to enjoy helping with that part too!)

Unfortunately, AI-written code was not up to this specialty task.

As another example, AI tools produced code in a way that would require annoying do-overs far too often.

AI code was not typically flexible enough for our uses, unless we took pains to instruct it over and over.

What do you really bring to your job? Probably many things…

After the AI finished working for us, we realized we were adding our own special ingredients, again and again.

These ingredients included:

• Specialized experience in the field
• A calibrated gut feeling for how things need to be done
• Knowledge of broader professional standards in working with tech

An AI probably only “knows” these things as if they are trivia questions. AI coding tools consistently failed to integrate them as necessary ingredients in the work.

In a similar way, most professionals who work with AI will probably need to add their own special experience, skills, or other attributes to the task at hand.

Oh, and the big picture…

AI tools also took instructions way too literally and could not see the big picture.

This is another one of our specialties. When a customer asks us to build a website, we constantly integrate and think about the big picture for them.

That “big picture” represents a lot of thought and practice, and includes everything that we know from working with teams, organizations, and websites.

AI tools don’t seem to integrate that kind of thinking very well, if at all.

Still, it’s a new trend. We’ll follow up to see what else we can learn in the future.

Conclusion

Here are some of our takeaways:

1. Don’t miss out on the AI wave! Try out the tools that interest you, and find out how they can help with your work.
2. AI tools will probably make many jobs easier and more valuable, contrary to fears. You can probably use AI to add value to your job.
3. It was kind of a relief to laugh at the new kinds of mistakes AI tools are making. AI tools are are often far from automatic and will fail even some very simple tasks.
4. AI tools do not accept responsibility in the same way your team members might expect you to. Be careful about trusting AI with your professional reputation.
5. There may be legal reasons to be careful with AI. It’s a good idea to be careful about publishing images you made using AI tools, for example. While some AI tools do not use any copyrighted artist data, others may not tell you whether they do or don’t.
6. It’s probably a good idea to be nuanced about AI, and not give in to the temptation to give a 100% good/100% bad verdict quite yet.
7. If you use AI in certain business or organizational circumstances, you should consider publishing an AI Usage Policy, or an AI AUP, or Acceptable Use Policy, for internal use.

Finally, one last point:

We could have written this article using AI tools. They certainly did some valuable writing for us as we tested them out.

But we didn’t use those tools to write this article. Instead, we wrote this article from scratch and revised it over thirty times.

This is a repeated theme in our work with AI: If you want to do professional work, you should still consider it a craft or an art that requires up-close refinement and hands-on awareness.

Well, until the great & terrible AI apocalypse finally comes…(maybe?) We look forward to learning more about AI, the associated tools, and their uses.