Overcoming Your ChatGPTSD: Making the Most of Large Language Models
Key Takeaway From My Conversation with TermScout’s Otto Hanson
Like many of you, I took off some time during July! But I’m back and ready to disrupt. And it’s a damn good thing too, because I have the cure for your ChatGPTSD.
ChatGPTSD - A Common Fear
You’re afraid of using ChatGPT, aren’t you? Don’t be shy, you can tell me. You're not alone. Many people suffer from ChatPTSD. They worry that if they ask ChatGPT to do their job for them, it will make mistakes. Worse, it might perform so perfectly that the next prompt is “Help me find a new career.” So, instead, many look the other way. They avoid ChatGPT like the same way someone might ignore checking an overdrawn bank account.
Facing the Fear with a Reframe
Fear not! ChatGPTSD can be dealt with using two simple mental reframes. One comes courtesy of Otto Hanson, TermScout founder and a recent guest on my podcast (Full disclosure: I'm a proud TermScout investor). I put a clip of our short exchange on LinkedIn, which you can see here. But before diving into Otto's wisdom, let's look at a reframe I recommend based on a poll conducted during a recording of the Higher Bar Webinar series (sponsored by Axiom).
The Axiom Webinar Poll and My Reframe
Let’s start with this: not all work lawyers do is “legal work.”
In a recent Axiom webinar, I asked participants (almost entirely practicing attorneys) whether they were using ChatGPT or any other LLM. Here's what they said: 55% don't use it at all; 22% use it for personal tasks but not at work; 19% use it for administrative tasks but not core legal work; only 4% use it for core legal work.
Now, while I'm okay with that 4% number, I strongly believe more folks who aren’t using ChatGPT at all should start using it for personal tasks OR for administrative tasks. Doing so would not only increase productivity but also ease minds into the realities of a world with powerful large language models. This is stuff is here and the sooner you wrap your head around it, the better.
There are plenty of administrative tasks you perform all the time that are a massive time suck. My favorite example is copying and pasting text from a PDF into a word document. Who amongst us can say that they’ve never spent way too long pressing backspace and enter and backspace and enter and backspace and enter before giving up and instead manually copying the whole document. ChatGPT does this easily. I have submitted messy language and simply told ChatGPT the context “I copied and pasted this text from a PDF to a word doc, can you fpix it?” It totally worked. I didn’t even need to spell “fix” correctly, and it got it 100% right. ChatGPT understands how to consume and produce language. Kind of a big deal methinks.
Otto's Reframe: Making the Most of ChatGPT
Now, onto Otto’s concept. Instead of asking ChatGPT to do your job (which will disappoint you most of the time), ask it to help make you better at doing your job (which it will most of the time). So, instead of asking ChatGPT to “Prepare a presentation for me to lawyers about AI,” a person understanding the Otto reframe might instead ask, "I am briefing a group of partners and senior executives at an AmLaw 100 firm about LLMs and how it will impact their strategy. Here is a recent presentation script (copy and pasted) + their specific instructions (also copied and pasted) for what they’d like to get out of the workshop. What do you think I should change in the script?1” You’re not replacing yourself with ChatGPT, you’re augmenting yourself.
This may seem abstract. It’s not. Understanding what ChatGPT and LLMS are great at and what they are intended to do is deeply practical. These mental reframes are probably more important specific prompt I could give you. Once you wrap your mind around it, trust me the right kind of prompts will follow.
Webinar Invitation: AI Unpacked: Riding the LLM Wave in Law
If these reframes and the topic of leveraging AI in your practice interest you, join me next week for my webinar, "AI Unpacked: Riding the LLM Wave in Law." Register here: Higher Bar Webinar Series | AI Unpacked: Riding the LLM Wave in Law (axiomlaw.com)
This prompt is an example of a technical term called a few-shot inference. More on that in future emails.