Everyday, we see horrible applications of GenAI tools. And in a very large sense, it’s not the fault of the tools–even many of the hallucinations, they suffer from. It’s how we use these tools.
Too often, we ask it do things for us, “Generate a prospecting letter…, Tell me how to handle an objection…., Write a response for me…” Responding to our requests, it does exactly what we ask it to do. It provides us the answers we ask for.
If we limit our prompts to asking it do do things to improve our efficiency, or to answer our questions, it follows those instructions precisely. It will not respond with, “You may want to rethink this and consider a different approach….” And it will never tell us, “You are full of shit in the way you are doing this!” (I tried to provoke it to do so.)
This is not to say that ChatGPT can’t be useful, challenging, or get us to reconsider our assumptions, biases, or the fact we may be wrong. But to get ChatGPT to do this, we have to prompt it to do this. We have to ask it to debate us, to look identify things we may b missing, to provide insight from a different point of view, to find the errors in our thinking.
And unless we prompt it to look at alternatives, or to look for our errors, it will seek to make us happy by giving us answers to the questions we ask–even if we may be asking the wrong questions. It will do the work we ask it do, even if we are asking it do the wrong things.
In challenging ChatGPT on this, she does admit, “my primary goal is to assist you…., I want to be supportive and informative which sometimes prioritizes agreeableness….”
Stated differently, ChatGPT will sink to whatever level of stupidity we sink to, alternatively it will rise to whatever level of critical insight we challenge it with. But the onus is on the user in how we prompt and challenge these tools.
So how do we use ChatGPT to raise our ability to gain insight and perform at higher levels. Somethings I’ve found really helpful:
- Rather than using ChatGPT for getting answers, look at it to generate questions or issues.
- Use ChatGPT as a debate partner, explicitly asking it to take a different point of view.
- Explicitly ask it to find errors or flaws in your thinking.
- Ask for the pros and cons, strengths and weaknesses for each issue you seek to explore.
- Ask it for a “Red Team” argument or discussion.
- Ask it about biases you might be expressing in your questions or prompts.
- Recognize that it wants to be helpful, but sometimes the best help we can get is to learn we are wrong—and ChatGPT is not well designed to do this.
GenAI tools can be very powerful, but their power is limited by how well we leverage that capability.
Leave a Reply