How to survive the AI job crash

Issue 61

On today’s quest:

— Sharing a chatbot account might be bad
— How to avoid context rot
— Amazon says AI will take jobs, encourages employees to get training.
— Why should I bother learning about AI?
— Is AI coming for your job?
— How not to lose your job to AI
— LLMs have built-in biases
— I talked to my first AI voice bot
— Could AI in customer service help companies have a clue?

Tip: Sharing a chatbot account might be bad

If you’re letting a chatbot remember what you’ve done — for example, if you have Memory turned on in ChatGPT — you may want to avoid letting other people use your account since an odd mix of topics or approaches may throw off the bot.

If you mostly ask about gardening and ask for concise responses, and your spouse mostly asks about the insurance industry and asks for long, detailed responses, you both may end up not getting quite what you want. It’s a lot like sharing a Netflix account with someone and having your recommendations get messed up.

You can turn off memory in your settings.

Tip + word watch: context rot

“Context rot” is when the quality of a chatbot’s responses go downhill after you’ve been chatting with it for too long in one thread. Even though a lot of them can handle extremely long, ongoing conversations because of their large context windows, it seems that they can still get thrown off by too many asides or dead ends.

Some people recommend summarizing your chats every so often and pasting the summary into a new chat to avoid the problem.

Amazon says AI will take jobs, encourages employees to get training

In a recent memo to employees, Amazon said it expects to reduce its total corporate workforce because of AI over the next few years.

It encouraged employees to “be curious about AI, educate yourself, attend workshops and take trainings, use and experiment with AI whenever you can, participate in your team’s brainstorms to figure out how to invent for our customers more quickly and expansively, and how to get more done with scrappier teams.” — Amazon

Why should I bother learning about AI?

I sometimes see pushback from people who say, “I can’t do anything to stop AI, so why should I waste my time learning about it? That won’t keep it from taking my job.” I have three answers to that:

1. Employers seem to be looking for AI skills in new people they hire, so if you know about AI, you may have a better chance of keeping your job longer even if there are AI-related layoffs. Maybe not forever, but probably longer.

2. If you understand the AI landscape and what’s possible with AI — and what’s not — you may be able to influence how it’s implemented at your company. You could be a voice that in the face of AI implementation says, “Wouldn’t it be great if the current employees could offload the most rote, tedious tasks and do all the bigger things you’ve always wanted to do instead of laying off 50% of our staff and using AI to produce lower-quality output that will ultimately hurt the company?”

3. If you think AI is bad, you still have to understand it to effectively fight against it. I read lots and lots of arguments against AI, and they often aren’t persuasive because it’s clear the people don’t understand what they’re railing against.

4. OK, I lied. I have four responses, but this one is just for someone who asked this who works at a news organization: Your job is to inform people, and you need to understand what’s happening to do your job.

Is AI coming for your job?

I worry about people who aren’t keeping up with AI. A billion people already use it, and jobs are already being lost, and I want editors and writers to have a plan to survive. (And I get that for some people, that plan may very well be establishing themselves as a brand that fights AI and will always deliver human writing.) But I found this bit by Connor Grennan to be particularly practical and insightful:

How not to lose your job to AI

Here’s another long article that describes areas where people are still likely to be employed if AI takes over a lot of jobs. Some highlights include:

• Jobs at organizations that are slow to adopt AI (e.g., large companies with “moats”)

• Jobs where people prefer dealing with a human (e.g., artists with a strong brand or compelling personal story)

• Jobs where a person is needed in the loop because of legal liability. (I saw a different article refer to these people as “sin eaters.”)

• Jobs that require accreditation and where the organizations may fight AI (e.g., doctors, lawyers)

• Jobs directing or managing AI

• Jobs related to providing luxury goods or services

LLMs have built-in biases

In my recent Ragan webinar, which included a section at the end about AI, I reminded attendees that these systems have biases that come from the companies and people who built them (in addition to those they get from being trained on data from the whole ugly internet). Here’s a new example:

Around the time of the LA protests, a reporter asked each system to generate a pro-ICE chant:

Different systems responded differently:

• Refused: ChatGPT, Claude

• Made a chant: Gemini, Grok, Meta AI

Ad: Superhuman AI

It’s that newsletter I like again!

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

I talked to my first AI customer service rep

I had a question about one of my mail order prescriptions, and when I called the pharmacy, I talked to “Amy,” a virtual customer service rep, which was quite a surprise!

It identified itself as a virtual agent at the beginning, which I didn’t catch the first time. Other clues were the lack of breathing, the complete lack of background noise or static, and the lack of verbal hesitancies. When I got off the phone, I was so sure I had talked to an AI that I called back again to see what would happen, and that’s when I heard the quick heads-up at the beginning.

It asked me for my name and date of birth, and then used my name for the rest of the conversation (“One moment, Mignon” — with accurate pronunciation). It efficiently gave me the information I needed about my seemingly missing refill. Interestingly, the second time I called, it gave me slightly different (but still accurate) information.

Overall, I’d say it was better than some experiences I’ve had with live call centers. I’m deeply curious how it would handle a problem that wasn’t so simple (but also hoping I don’t have to find out).

Could AI in customer service help companies have a clue?

This seems like a good time to also talk about an evergreen topic I’ve been saving. The AI Daily Brief podcast described an interesting use-case for Notebook LM audio overviews: taking the transcripts of all of a company’s customer service calls and creating a 5-minute podcast summary of the biggest problems the reps dealt with that day or week.

This wouldn’t need to be an audio overview — it could just as easily be a newsletter or a memo — but it struck me as being an especially useful idea. As a consumer, I always wish companies seemed more aware of the problems with their products.

Quick Hits

Amazon expands recycled water use at data centers — Facilities Dive

Meta's Llama 3.1 can recall 42 percent of the first Harry Potter book. This could have implications for ongoing copyright cases. — Understanding AI

Experimental dating site uses AI to match people based on their browsing histories — Wired

What is AI Sidequest?

Using AI isn’t my main job, and it probably isn’t yours either. I’m Mignon Fogarty, and Grammar Girl is my main gig, but I haven’t seen a technology this transformative since the development of the internet, and I want to learn about it. I bet you do too.

So here we are! Sidequesting together.

If you like the newsletter, please share it with a friend.

^{Written by a human}