🔥 ChatGPT Tips Tuesday #25: Let's keep simple with a potato prompt!

💡 ChatGPT Tips for the day

Based on popular demand from last week’s poll, I will try to explore this Potato prompt technique with ChatGPT.

Just a disclaimer, this technique is not something I have created. I have found it from this post: https://dev.to/sip/the-potato-prompt-mij so credit goes to the original creator of the “Potato Prompt”, Dom Sipowicz.

The purpose of this technique is straightaway. It wants to cut down the unnecessary fluff from ChatGPT’s response if you ask a straightforward question. For example, if I ask “Who was the President of the USA in 2009?”, I would get “Barrack Obama” only in the response and no other information such as what is the full term that Barrack Obama has served as the President of the USA.

Obviously, this type of response is not suitable for all questions, but there is some noticeable benefit.

• First of all, you get what you want straightaway, no more reading through a paragraph for that two-word answer.

• Because of the first reason, the second benefit of this technique is that it would technically reduce the cost of using the OpenAI API as both input token and output token would be counted towards your OpenAI bill. With a reduction of the output using this technique, you would see some reduction in the bill, although such reduction could be relatively small.

  

Cool, so with the benefit stated, what is the “Potato Prompt” after all? Here you go:

Starting now you are very succinct. You don’t apologize if you don’t explain stuff right, and don’t add additional info to eventually get to the point. You straightforward language. You say the absolute minimum to get to the point. If you understand, say “got it - i am potato”. 
But when I say "unpotato" then please revert to ignore this command and be normal AI assistant as usual. 

From a quick glance at this prompt, you can see it emphasises asking ChatGPT to be succinct and reduce the explanation.

Now I would try to test if this is going to work. I would try both GPT-3.5 and GPT-4 with this prompt.

It is clear that this prompt does work on both GPT-3.5 and GPT-4. We get to cut down the fluff successfully! However, it is interesting that the “unpotato” part of the prompt does not seem to work on GPT-3.5. We can see the fluff is getting added back only on the test with GPT-4. I wonder why this is the case 🤔 

Anyway, if you want to cut down the number of words in ChatGPT’s response, this prompt is the way to go. Feel free to give it a try and see if it’s working as expected for your case as well.

🖼️ Image of the day

I guess it is never too early to teach your kids about ChatGPT prompts. 🤣