How I'm using AI right now
but not like, RIGHT now
I know a lot of folks who reject AI in every form, and while I have thoughts about the limits of this skepticism, what I’ll say now is that familiarizing myself with these tools has been important to my work, because my clients are using them no matter what. I also know that a lot of people are finding that their favorite platforms suddenly have new “helpful” features that they’re not sure how to feel about (or opt out of). I recently decided to pay for ChatGPT, and so I’ve been using it pretty freely for a couple of months. Here’s what I’ve figured out about how it fits into my life and where it doesn’t.
What I use AI for:
At a high level, the goals of my AI use are to supplement and extend my executive function, help me solve problems I run into in the course of my work, and saving my brain a few cycles on things that are in the “overhead” category of things you should actively try to do badly (h/t to Shreyas Doshi, one of the only worthwhile product thought leaders, for this framing).
Being a half-decent brain on a bad day - Most days, I can write a better email than ChatGPT can. However, if I have a bad head cold and need to write a high stakes professional email, AI is a lifesaver. Emails, as a rule, don’t demand 100% originality, you just need the words to get the correct message across.
Helping to structure writing - Maybe it’s my unmasking era, maybe it’s that I haven’t written much in non-work contexts for years, but writing isn’t the way it used to be for me - a lot of what I write down is just jumbled stream of consciousness that contains the general nugget of what I’m looking to express. Rather than ask ChatGPT to write the post for me, I’ll give it what I’ve already written and ask it for a suggested outline I can follow to put everything I’ve already written into a more sensible format. There’s also an accessibility angle here - when I was in the throes of my wrist issues, I used AI to take voice-to-text blocks and format them into posts. This is how I discovered that I don’t like what AI does to my written voice, but it at least gave me something to edit (much less hand intensive) rather than requiring me to type everything myself.
Polishing writing - While I don’t like the way that ChatGPT tends to rewrite things, I do appreciate a little editorial polish. I’ve found that you can add directions to queries that keep your original voice intact while editing for clarity, fixing weird punctuation, and so on. As I discuss in the next section, I still edit anything ChatGPT polishes for me (also because it always ends up sounding more girlboss-y than I ever intend to).
Perseverating - I need a lot of reassurance, which it turns out AI is great for! Rather than ambiently blast my group chats with thoughts like “is it okay to be 4 minutes late to the vet” I can dump some of my anxieties into the chat box and get the comfort I so deeply crave. I have ambivalence about the interplay between AI chatbots and the ideals of individualism, and I don’t think we should all be getting our support from bots, but for those little moments when I am feeling some type of way about something silly, it’s helpful.
Answering complex queries that need to take multiple factors into account - This is good for both work and play. For some of my projects, I need specific technical advice given a set of parameters (variables in these formats, etc). I could spend hours piecing together the information I need by combing forums, but typically ChatGPT gets me most of the way there pretty quickly, and I can supplement what I don’t know with further research on non-AI platforms. On the play front, sometimes I want a list of recommendations for movies, books, or activities that fit very specific criteria - when I had a Twitter account with thousands of followers and a broad reach, I would use that, but since I withdrew from that world it’s good to have a source for “10 offbeat action movie recommendations with at least a 120 minute runtime and no subtitles” for playing in the background while I hang out with the rats.
My rules for using AI:
I may write this up into an actual AI disclosure someday (see Katie Harbath’s Anchor Change newsletter for a great example of what this looks like), but right now that feels grandiose given the scope of my newsletter and the fact that I’m not taking sponsors or paid subscriptions. Rather, this is my personal code to keep my usage of AI up to my ethical standards, as it’s developed over time.
I don’t put real data into my queries - eg. if I am asking a question about how to do a data thing for a client, I supply hypothetical variables. I anonymize situations that I ask about, removing names and identifying details.
I edit and research liberally - I don’t copy/paste things from ChatGPT and then leave them completely unedited - I always go in and make sure my voice is there, even if it’s as simple as putting my typical signoff on an email. If I’m taking a phrasing that ChatGPT has produced, I am checking it against other sources to make sure it’s not plagiarized. I also don’t ask ChatGPT for new pieces of writing or facts from whole cloth, because I don’t trust that it will be correct, and also because I’m a freak who enjoys research.
I stay away from generating media - I tinkered with the image-generation tools like Dall-E when they first came out, and my LinkedIn photo is still an AI-generated headshot that desperately needs replacing (there will be a post on that once I get new headshots taken). I have also used these tools to generate icons for my business, but my plan is to replace those with logos that I pay someone to create once I’ve hit a couple more profit milestones. I’ve read enough about the environmental impact of media generation, as well as the impact of AI-generated art on creatives, that it makes me want to stay away from it as much as possible. I’m not asking it to generate images just to amuse me, or to fill space in my newsletter. Y’all will just have to deal with my bad photography, cringey memes, or whatever other visual interest I can find on the honest internet.
I don’t revise endlessly - I’m not asking ChatGPT to re-generate things over and over until they come out exactly right, both because that’s time-consuming and computationally expensive, and also because it removes a lot of my voice from the writing. If AI gives me something that’s even 50% of the way to what I want, I will take that and make my own updates.
What I think we need to make AI more sustainable:
Let people opt out of AI “help” - eg. the now-everpresent AI blurb in Google searches. Some folks have found workarounds for this, such as searching for “fucking broccoli” instead of “broccoli,” but I’m curious about whether that means the AI-generated prompt isn’t being created or just isn’t being shown. In any case, given the environmental impact of AI, I think that consumers should be able to choose when AI gets used without needing to get potty-mouthed with their queries.
Don’t offload important decisions entirely to AI - I’m not seeking financial or medical advice from ChatGPT. If I had the power to impose tariffs, I sure as hell would not get them from there either. If a business is making decisions using AI, such as deciding which content to moderate, a human needs to be somewhere in the loop.
Media literacy and critical thinking - This is the #1 solution to mindless AI hype, but unfortunately the one that’s most difficult to get people to actually do. Remember that AI is just a tool, and you need to critically evaluate not just its output, but also whether to use it for a given task at all.
As an industry, invest deeply in understanding AI safety - Just like with so much technology, AI has been deployed in many cases before we’ve been able to really understand the dangerous impacts it can have. While the AI hype train is in motion, the tech industry needs to invest more in understanding the risks of AI and preventing them (which is why I do appreciate some of the AI companies like Anthropic taking on a more research-focused angle).
Ultimately, one’s relationship with a new technology is deeply personal. This is just where it lands for me, but if you haven’t experimented much with AI tools yet, then maybe this can be a helpful guide for you as well. If you think I’m depraved for using AI at all, that’s fine too.