I forced myself to sign up for all sorts of premium subscriptions which I don’t usually do, because I’m so likely to forget to cancel the subscription before the trial ends. Of course, AI can remind me to cancel the trial, but AI can’t stop my habit of saying, “Oh yeah. Okay I’ll do that as soon as I finish what I’m doing.”
So anyway, here is the list of AI I paid for in the last month and how I used it.
ChatGPT – The best way to learn totally new technology is to see how other people use it. This is how I learned to use the Internet in 1993. For example, I saw someone talking to a boyfriend asynchronously and I wanted to do that too. So I learned prompts before there were browsers. After using new technology for a while, we’re more likely to dream up our own uses. For example, I learned HTML to publish stories I wanted people to read out of sequence. With ChatGPT I told myself to trust this process, and try something new every day.
Claude – My brother at [fancy university] told me professors there are using Claude for writing and ChatGPT for research. So I tried to divide up my tasks that way. My friend at [fancy law firm] said when they receive a motion they ask ChatGPT to write a response, and they often just make small edits. I hear this and I have AI FOMO: I need to learn faster. I discovered that the less I know about what I am trying to write, the more likely it is that ChatGPT will do better than Claude.
Beautiful.ai – I tried this one because I love making slide presentations, but I got scared I was nurturing a skill that AI could do better. I tried some ideas with Beautiful.ai and the operative word here is beautiful. The slides look really good, but when I uploaded the text for the slides, the AI rewrote it. Not helpful to me. So I haven’t found a good way to use this one: comforting, in a way.
Canva – I’m good at using this software, and I noticed that when I get good at using a piece of software I don’t even realize when I’m using AI because it’s baked into how my brain is leveraging the tool. With Canva I have a good sense when I’ve hit a wall with ideas and I need to ask for help. For example if I can’t figure out how to lay out information on a slide, Canva gives a bunch of suggestions. And, even if they are not what I ultimately use, I learn a little bit more about thinking visually.
Grammarly – If you think of yourself as the manager and AI as your employee, then it’s not surprising that the easiest AI to manage is the type that’s great at what you’re great at – you know when they’re doing a good job and when they need help. I love Grammarly because I love grammar. Sure, Grammarly makes some annoying stylistic suggestions. And it hates if you start a sentence with and. But every time I ask Grammarly to proofread I learn a new way to write more clearly.
Dall-e – When I asked Dall-e to fix the lighting in my photo of my dog, I got the photo above with the explanation: “I highlighted the tail.” Except my dog doesn’t have a tail. Using Dall-e reminds me of managing a rock star at work; they can do great stuff but it’s difficult to get them to follow directions that would make them useful, so you mostly just wait for them to surprise you.
Copilot – I read that Copilot does the work of a manager for developers. So I thought, what would I do better if someone were managing me? (And I thought: probably everything since I’m so erratic.) I tried a book proposal. I have 50% done but I need to finish the rest. I gave the job of managing my output to Copilot and I got a project plan based on the numbers I plugged in. When I gave the job to Claude I got back a less detailed plan, but Claude initiated a discussion to fill in more details based on my particular concept.
My conclusion? I need to push myself to think in terms of the highest common denominator. Claude taught me that before I ask AI to manage my book project, I should ask AI to help me figure out what the book project is. But wait. Can AI write the whole book? I don’t know. I’m not alone in my cognitive dissonance about asking AI to do something I feel especially good at. Doctors, for example, ask AI for help thinking about how to diagnose (e.g. what are some avenues to investigate for x symptom?). But AI is better than doctors for diagnosis itself (e.g. here are the symptoms x, y, z what is the diagnosis?)
This whole process reminds me of a classmate telling me the most important part of college is learning how to ask better questions. If colleges want to be relevant again, they should really lean into the idea that they can teach kids how to ask better questions for AI. Because that’s going to be everyone’s job in the future.