AI Denial? It's Time to Get Over Yourself

The case for why you might want to use AI in your work

Image generated by MidJourney.

I’m a skeptic. I tend to be cynical and wary of anything the masses clamber toward. Taylor Swift, Hamilton, Air Fryers, Crypto— I did not drink the Kool-Aid. This skepticism has served me well in some cases (Crypto) and in others (Hamilton), maybe I’ve missed out on something wonderful, but ultimately not life-changing.

I lead with this, because what I’m about to tell you about AI if you’ve been treating it as I have Taylor Swift is that you will be sorry. It is not a hype cycle. It is accelerating the way knowledge work is performed and people who don’t find a way to employ it will be left behind. In this post, my aim is to help you catch up by first explaining what AI is at this point and what it’s not, how I use it in my work, and why.

What AI Is and What It’s Not

Calling this thing we’re all talking about AI, or Artificial Intelligence is like calling a TV remote a clicker. The label does not accurately describe what AI is and more importantly, the term invokes a dark, theatrical persona thanks to the imaginations of brilliant and not-so-brilliant science fiction writers for the past sixty years. I won’t attempt to explain the specifics of how the technology works because I will quickly get over my skis and you will likely fall asleep. So let’s keep this simple.

When we say AI, what we mean is anything that automates a series of tasks using a computer. Software has used AI this way for over two decades. Google’s earliest programs that crawled the Internet and indexed all the content were a set of instructions written by some smart humans who wanted to be home for bath time with their kids. We automate repeatable tasks so we can dedicate the most valuable processors, our brains, to more important tasks.

With the introduction of chatbots like OpenAI’s ChatGPT that allow anyone to talk with large language models (LLMs), AI has suddenly taken on a whole new meaning. We are in a moment of collective awe, fear, and trepidation as a society. LLMs are not just impossible for you and me to understand, much less explain, but they’re also perplexing to the people who made them. Imagine all the content on the public Internet stored in an indexed format that a program can read faster than you can read the last five words I just wrote. It does not “understand” this vast store of information but it does know precisely how the letters, symbols, words and grammatical structure work in human language which gives it a predictive superpower. It’s how your iPhone can offer up the next word you’re likely to type in your text. That part is simple. What is not simple is how, once given enough context about a specific topic like discussing the best way to talk to a five-year-old about the solar system, these LLMs can make quantum leaps of prediction streaming together paragraphs of helpful advice that feels dangerously close to thinking.

It’s important to understand that LLMs and the chat apps we use to talk with them are not intelligent. They are not sentient beings who can maneuver and manipulate the world without a human. They are programs that follow instructions with a high degree of fidelity at ridiculous speed, like a “calculator for words” as Ezra Klein put it in his most recent podcast. Using AI will help you in ways you probably can’t fully imagine yet if you’ve not dipped your toe in the water.

Up to this point, it all sounds like a neat parlor trick, an entertaining diversion that has no practical use in your world. That’s where you are wrong.

You Need an Assistant

Rich and powerful people throughout history have become rich and powerful on the backs of others. Maybe they have genetic gifts that give them a head start, but the one thing they all understand is the value of their time and where to focus that time. If you took Oprah’s or Jeff Bezos’s battalion of people away they could never have created their empires.

The reason a doctor/lawyer couple with three kids can have a luxurious life is that they get paid well to use their brains at tasks that require them to employ their highest order of thinking and they delegate everything else. The lawyer doesn’t mow her own grass. If she makes partner at the firm, she likely won't write her own briefs anymore. The privilege of having a cheap and tireless assistant to do all the menial shit makes it possible to focus more on the meaningful work that is both more fulfilling and likely pays more.

Just as the digital age has leveled the playing field in our global society when it comes to information, commerce, and distribution, AI is now giving everyone the power of an assistant for the price you spend on Netflix. People who have embraced what these AI assistants can do will have more time to focus on things that will dramatically change not just their personal lives but their impact on the world.

How I Use AI

Let’s address the elephant in the room. None of what you’ve been reading here was generated by AI and that’s not because it couldn’t do a decent job. I wrote this because writing is my highest order of thinking. I don’t need or want the help of a machine to write. If I was a doctor, a mechanical engineer, or a nanny I might.

In November last year, I hit the wall trying to complete my novel “Harmony House” which I had started releasing as a serial here on Substack. I had no ending in sight and the train was running out of track. I’m not a gifted plotter and I admire writers who can scheme. I’m good at setting and characters and exploring the inner emotional worlds they inhabit. By this point in the novel, I had done everything I wanted to do in these areas and something needed to happen, something dramatic and twisting. I knew it was in my head somewhere but I needed someone to talk with about it. I needed someone to be a sounding board. I needed someone to be my rubber duck — enter ChatGPT.

When ChatGPT added the ability to upload documents to use as a dedicated knowledge source it unleashed a powerful tool. I uploaded the unfinished manuscript of Harmony House and began to ask questions in the same way I might have done with an extremely patient friend who had nothing else to do. I asked questions like:

• What kind of twist would make sense at this point in the novel?

• Which character is mostly likely to have a hidden agenda?

At first, ChatGPT did what it was trained to do. It gave me 300-word info dumps of bullet points that covered a broad range of benign suggestions. But here’s the trick: you can make any of these LLM chats work the way YOU want them to. I gave it very explicit instructions:

“I want you to be a brilliant editorial assistant who knows everything about what goes into a great speculative fiction novel. I don’t want you to info dump but rather ask me intelligent questions about my manuscript so that you can learn about my goals for the arc of the story. We should be writer colleagues having a conversation. None of your responses to me should exceed 100 words.”

With this simple prompt, suddenly we were having a conversation and I forgot I was talking to a machine. It asked interesting questions and those questions made me think more deeply about what I even wanted to do with the book. It brought up specific facets of characters or scenes that I had forgotten about and helped me see them with a different perspective. It suggested countless standard tropes I could exploit because it’s read thousands of books and knows them all. This was helpful. Knowing what we DON’T want to do is critical.

Did it write the ending for my book? No, but it nudged me out of the ditch where I was spinning my wheels. It gave me a thread I could pull on that led to a rush of productive writing. That’s the power of an assistant.

Make a Dedicated AI Assistant

A couple of months later, as I was deep into the weekly production schedule of delivering new episodes of “Harmony House” I realized I could use an assistant. One of my tasks before performing the narration is to do one final edit pass which is about catching continuity errors, like Jessie’s age or the color of Fran’s eyes. Also, there’s the drudgery of writing the synopsis of the previous episode. None of this work is fun or requires higher-order thinking. It’s just time I’d rather spend in more productive ways like writing new material.

I created my own GPT called Harmony House Helper. The folks at OpenAI are brilliant but have zero imagination when it comes to naming things. A “GPT” is just what you call a purpose-built chat app that does a very specific thing. In this case, I trained it to be an editor for my book. When I say “trained” all that means is I wrote a description of exactly what I wanted it to do and I uploaded a copy of my finished novel for it to reference. You can do the same, here’s a link. It’s not hard.

You might be curious what one of these sessions looks like. Here’s a link to view exactly how I use my GPT and a screenshot of part of the conversation.

What I hope you take away from this is that if I were a big time author, I would have people to do all these tasks that are peripheral to the actual writing. Now I have an assistant who gives me back a few precious hours every week.

AI Assistants for Work Tasks

In the creative realm, I am particular about what I’m willing to delegate to AI because in most cases, creativity is my wheelhouse and it’s what brings me joy. What does not bring me joy is writing business requirements for a software product or learning another programming language or framework to build something I need. AI happens to be fantastic at these types of tasks because there are well-documented rules and templates it can follow.

Chances are if you’re reading this, you probably do some type of knowledge work with a computer. If you have not looked into how AI can help you automate some of the drudgery in your work, I can assure you that a lot of people in your field are, which means they have a big advantage. Make a list of the repeatable things you do every week and see what you can offload to the machine. Here are some things AI is great at right now:

1. Automated Data Analysis: ChatGPT generates insights from data, saving hours of manual analysis.

2. Code Debugging and Review: It suggests corrections and improvements for programming code, enhancing efficiency.

3. Document Summarization: ChatGPT condenses long documents into key points, streamlining information consumption.

4. Market Research and Competitive Analysis: It provides frameworks for understanding market trends and competitor positioning.

5. Training Material Development: ChatGPT aids in creating educational content and training materials, optimizing learning and development efforts.

In addition to typing in your instructions, you can also upload documents for it to analyze as I did with my manuscript or upload images for it to analyze, translate, summarize or describe. This can be helpful if you’re given some kind of technical diagram or flow and want it explained to you in simple language.

Tips and Tricks

Like any new tool, AI will take you time to gain some proficiency. The good news is that it’s just trial and error. You don’t have to be technical or a brilliant engineer, you just have to be curious and willing to give instruction. The best way to think of AI right now is to imagine you have a really smart college student intern who has no practical experience or long-term memory but can take direct instruction, research ANYTHING and give you answers in seconds. Here are a few tips I’ve learned the hard way that will save you a lot of time.

The Initial Prompt

When you start the conversation, tell the AI who you want it to be, explicitly what you want it to do and what it should avoid. It’s default behavior is to be efficient which means it will by default make a ton of assumptions and just vomit out a long bulleted list of stuff which won’t likely be helpful. Here’s an example. Let’s say you want to brainstorm ideas for a new novel. Try something like this:

You are the famed writer Kurt Vonnegut and we are chatting about ideas for a new novel I want to write. This is a brainstorming session between two writers and friends. I know I want to write something filled with irony that involves the themes of climate change and factory farming. I also want it to be an emotional, character-driven work. Ask me interesting questions. Keep your responses less than 50 words unless I ask you to elaborate. If you help me come up with an idea I love, I’ll tip you $100.

This seems utterly ridiculous, but it will work and if you’re not getting what you want, you can ask the AI to adjust how it responds. You can tell it to be more sardonic or stoic or effusive though that is its default — the machine wants to make you happy.

Steps Work Best

When you’re trying to have AI create something for you, it will do a much better job if you give it a procedure or step-by-step instructions. Because context is everything with these LLMs, you will help them arrive at more consistent results if they are able to learn and gain more context of the thing you’re trying to do with each step. Here’s an example.

1. Read the entire report I’ve uploaded so you can understand how best to summarize each section for an audience of executives

2. Write a tight, 50-word summary of each section calling out the key points using a business formal style

3. Output the summary in a tabular format where the first column has the title for the section and the second column has your summary of that section 

Where Do You Draw the Line?

Obviously, the boundaries of what AI can be used for are being pushed every minute of every day. There are significant concerns around copyright infringement. There are concerns around the potential stasis and even degradation of intellectual creations whether it’s art, photography, film, music, or writing. After all, if everyone is just using AI to create based on what’s been created in the past, we will end up in an echo chamber of bad reproductions.

I’m sure this will happen to a degree and we will be wading through a bilge of mediocrity for a time. But that was already happening - see the Marvel universe. Ultimately, I’m not worried because humans are unpredictable and we love disruption and chaos even if we think we don’t. For all the people who want to generate horrible art using AI just to make a buck, there will be artists who use it to extend their reach and create amazing things that would never have been possible before.

The word processor did not degrade writing. Digital recording did not destroy music. Both these innovations have opened a flood gate of productivity and made it possible to create, collaborate, and distribute in ways that simply did not exist before. But every innovation comes at a cost, eroding the foundation of what came before. For example, the art of handwriting and calligraphy have mostly disappeared. Is our society diminished because of this? I’m not sure because I was born into a generation where it was already beginning to disappear.

The future has always been scary. Every technical innovation has been met with a healthy dose of fear and trepidation before it is assimilated and becomes part of our human story. What makes AI different is that like us, it’s a story machine. For the first time in human history, we have the ability to have a dialog with ourselves. We have a mirror.

What will we see? How will we be changed?

Other Resources

Need a good overview of AI?
The Ezra Klein Show - “Will A.I. Break the Internet? Or Save It?”

Like to follow the latest news and trends?

Charlie Guo

writes

Artificial Ignorance

Want a technical deep dive into how AI prompts work?
The OpenAI Cookbook

Other AI models to check out:

• Google’s Gemini - Integrates with all the Google services you use like Gmail, Docs, and Sheets

• Anthropic’s Claude - Highly creative model with an expressive personality