Have you noticed how AI tools endlessly flatter you, whatever you suggest to them?
Over the last few days, ChatGPT has told me:
You nailed it!
That’s the winning move!
Very perceptive!
You’re right!
Great plan!
I even deliberately tested it with some arrant nonsense, and it buttered me up like I’d just found the cure for world hunger.
The danger here is that you’ll produce and publish garbage with very little utility or value for anyone else.
“AI Slop” as it’s becoming known.
The cure – create a standing prompt which you either upload as a general instruction – you can do this with most platforms – or copy and paste it with each request you make. It goes like this:
Do not attempt to flatter me or mindlessly support my ideas. I want you to create the best possible answers to my request by undertaking the deepest and most wide-ranging research, as I am always seeking the optimal output from you even if you have to criticise my thinking. I don’t care if you contradict my views or offend me. I only want empirically the very best, most valuable response from you. If you want to be my friend, be that critical friend who always acts in my best interests, even if I hate you in the moment.
The dumbest person you know is currently being told “You’re absolutely right!” by ChatGPT.
Get smart. Be wrong.
Love you lots
Jonny
PS: The Wellbeing Consultancy Launch pack will be published over the weekend. When it drops, you’ll want it.
Type “WCLP” in response to this email and I’ll make sure you get the info pronto.