Working With AI: Like Training a Brilliant but Flaky Intern

I listened to a podcast featuring AI & Tech sceptic  Ed Zitron, he made some great points and although I do like AI and use it,  I can’t help but find something unnerving about the way it speaks.
 
I’ve seen it produce impressive work in minutes and at Modern Citizens we’ve  been carefully looking at how to integrate AI into our workflows ethically  and in the most optimal ways. Advertising (and Branding) often thrives on the  hard work of freelancers and external suppliers at every step, the human  touch at each stage is what makes a campaign successful and also unique. You  are getting a reaction and approach based on someone’s lived experiences over 20,30,40,50 years. Hook it to my veins! I could wax lyrical about how no two approaches  will ever be the same. It's basically the same thing as every time a deck of  cards is shuffled it’s apparently never been in that order before (which to  my pathetic brain sounds scary and impossible – but I’m only human).
 
More recently I was tasked with creating some pre-production scamps on a  tight deadline and I thought it was a perfect opportunity to explore Chat GPT’s visual image capabilities.  
 
After some excruciatingly pleasant greetings and chit chat (as well as some  top tier brown nosing on the creative idea) Chat GPT got under way and  produced something very mid. I’m sure my prompts were less than ‘optimal’ but  I’m only human. With some refinement we managed to get to a place that would  make great foundations for me to start comping on top of and generally clean  up. We were a team, like 2000’s era Edge & Christian, GPT’s pleasantries  become more tolerable. I found myself warming to them. After some final check  in’s off they went to create the final version.

🤖 “I’ll need 90 minutes  to create this”
GREAT! See you then pal!
 
I checked in after 2 hours, with a polite request on an update.
‍
🤖 “I'm about halfway through, I’ll be done  in 30 minutes”

For an all knowing, supposedly nearly sentient, industry changing piece of  technology it had a rather lacklustre attitude towards time management.  
I let it slide and checked in again an hour later.

🤖 “ about 15 mins left”
Now I’m becoming impatient, the prompt channel was a long stream of passive  aggressive ‘how long now’.
 
Hours after the original deadline, I received the message I’d been waiting  for
🤖 “Uploading it for you here:”
but nothing materialised, “I can’t see any image”
 
🤖 “I’ll add a download link”  
I clicked the link to receive a “file not found” They then  proceeded to completely forget about any previous conversation and just  pretend it never happened (some nuclear level brown nosing on the creative  idea now ensued after re-briefing).
 
I foolishly proceeded to endure this cycle 3 or 4 times because…hey, I’m only  human!
 
In short, the final result was a great collaboration, Chat GPT did the hard  labour. A talented but wayward young trainee I took under my wing and moulded.  Reigning them in when they go rogue or turned up late after the Christmas  Party. SO after the first foray into a working relationship with Chat GPT – here’s  5 top tips I would use in future to ensure a smoother, more OPTIMAL result!:
 
 🗣️- Use short ANGRY  commands (preferably in CAPSLOCK.)  
 💦- Pour water on your keyboard to ensure maximum hydration!
 📈  - Threaten them with a performance plan.
 💸  - Steal their lunch money.
 🐕  - Let your dog mess on their doorstep.
 

Drop us a line if you would like to see more of what we do when we aren’t wrestling with Ai!