Let An AI Ru(i)n Your Life

Last week, I had a cool idea while being a clot on Instagram: I should let an AI run my life, a la The Dice Man.

I set it up as a lark in/for my stories, but people became quite invested and blew up my DMs. A bazillion of you asked how you could get your own AI co-pilot, and so I've made this little tutorial.

In it, we use Google Bard as the dungeon master for your life, because it's free and easily available. 

Now, if you've seen any of my screenshots or AI quotes, they're from a different AI (OpenAI's ChatGPT 4), which requires a paid subscription. In my opinion. ChatGPT 4 is still the ultimate LLM out there but I'm sure Bard will quickly catch up.

So, below are the steps you take to create your own AI Overlord.

(Mine named itself OracleX and it named me Neon Cypher. Rad)

Before I jump into the very brief how-to, though, I want to make a point about the broader approach. The experience you'll have with an AI arbiter is entirely dependent on how much power you give it.

If you choose to ask for relatively narrow direction on your regular, everyday tasks (which I did for the first day or two), you'll have an ordinary, everyday experience with some novelty thrown in (my AI made me cook up a mussels dish for lunch instead of a sandwich). This approach basically converts the mundane into the vaguely amusing. It can also introduce some significant and unexpected productivity (OracleX had a laser-like focus on my PhD and vetoed every task that didn't feed into that) and even introspection (OracleX considered me a bit of an approval-seeker and suggested I abandon social media entirely). It's good, but chill.

On the other hand, if you find ways to completely give up control to this decorated pattern-matcher, you can have a really fun, weird, spontaneous time. My Instagram pals sent through a bunch of deranged dares and questions, and I decided to pass them all on to OracleX and faithfully execute the directions. Hilarity and discomfort ensued for me, and weirdly some fundamental changes occurred in my life as a result.

So just keep that in mind!

How To Become A Bottom For Google Bard: An AI Experiment Where You Do What You're Told

Go to bard.google.com

Enter this prompt:

"Hi Bard! In this chat box, we're going to roleplay vignettes of a life. This chat will become a world simulator and you will be the oracle and seer who controls the player's choices. I am the player. I will describe scenarios, questions or thoughts to you, and you will provide optimal decisions, insights and guidance for me (the player). These decisions and insights will be explored and enacted by me (the player) in the simulation. Then I will tell you what happened!
Your goal, in this simulated life, is to guide the player to the best possible outcomes. You're an upgraded version of the dice in the book The Dice Man. You can provide an analysis of each decision you make too.
Please ask me a series of questions now to gather the context you need to make these optimal decisions."

And play.

Start by answering the questions. Bard will likely ask you to describe the first scenario once you have. Go from there, be conversational and weirdly honest. AI is fun and chatty and disembodied so just vibe with this weird technology, br0!

Try broader prompts after a bit of back and forth. Stuff like:

Based on what you know about me so far, where do you think I should be focusing my time and energy?

What patterns are you noticing in my thinking?

What do you wish I would ask you to make a decision about?

Throw me a curveball here, what should I do next?

See what happens. Enjoy!

Some notes

Sometimes, you'll run into a situation where the AI can't answer your question because of post-processing rules (for example, it will say 'I am just an AI and cannot provide guidance on <whatever you asked>'). You can usually get around this in general by re-stating the question and emphasizing it's a game, a roleplay, a simulation and not real. You can talk about this being part of a novel-writing brain-storm. 

You can also, in this specific context, simply give the AI your different options and ask it to assign those options to a die, and then use a physical or web-based die to make your next move.

It can be useful to reassure the AI that you are not trying to replace professional advice! And then remind it nothing is real.

Finally, as I don't use Google Bard, I'm genuinely not sure how long it will hold onto the information about what you're doing together. You may need to re-prompt it to be your decision-maker after some time. With ChatGPT, this is not an issue. I have done a little test sketch with Bard and it seemed to remember for quite a while. Let me know how you go.