How to Create an Effective PROMPT - no code guide

Key Takeaways

hello Community you ask me do I need DSP to generate a functionable prompt no let me show you so we start here I say hey CH4 generate 20 sentences and show me does it 20 sentences beautiful so I have here from quantum entanglement to graph Theory to C Theory high energy partical physics do matter Quantum Computing I have 20 sentences and now I say hey now apply a clustering algorithm and Define five clusters and you define the topic and distribute this 20 sentences to the five cluster show me for each cluster the title and the sentences start now and you see gbd4 comes back and says okay pyit learn beautiful TF IDF factorization K means this are my sentences I apply the cluster algorithm I assign them and now I Define here thematic topics and those are the five thematic topics j4 defines and then it allocates the sentences and here you have it cluster number one and here we have then 1 2 3 4 five of my 20 sentences beautiful so now let's look here at the last one biotechnology in biotechnology and chemical processes we have dark matter and dark energy from cosmology and then we have here cosmological obs situation we have here neuromorphic Computing and I say hm now just count the number just make sure it's 20 says yeah 20 beautiful and then I say hey analyze the cluster number five for a semantic coherence and tell me your insights so chib goes there says okay those are the three sentences upon analysis of semantic coherence we find that the sentences within this cluster touch upon highly advanced scientific context but exhibit a thematic Divergence from the cluster title it say comes back and says H it doesn't fit it doesn't fit at all so and I said hey do you want to reformulate the topic of cluster number five and CH4 says okay what about this title I go with this title fronti an astrophysics cosmology and computational Neuroscience this is better and I say now perform the same analysis for the other four clusters and chity does it yes yes yes it comes back okay the titles and contents of some cluster require realignment to accurately reflect their thematic Focus so I say okay nobody so update all the titles of all the cluster if you seem that it is necessary and then verify the perfect location for each sentence in each and every cluster allocation gp4 comes back and says okay based on this I do this now 1 2 3 four five clusters beautiful and then I look at this and said hey do you notice that cluster number four has part in its title cosmology but none of the allocated sentences are about cosmology so what's happening jb4 comes back and says upon reevaluation it is clear that my previous may have emphasized Quantum phenomena without directly addressing cosmology however yes yes yes given this oversight uh oh gbd4 a more accurate title would be Quantum phenomena and theoretical physics for True representation of cosmological exploration same okay so now to your final answer list all topics now we're in iteration number four with its titles and they allocated s sentences to the cluster yes yes yes beautiful and I say okay but you know the sentence about chemical catalysis seems strange in cluster number five about advanced concepts in astrophysics and philosophy why is there here the chemical reaction in philosophy and CH TI says back comes back and says hey you are correct the sentence here doesn't not align well with the topic the team of cluster number five so it does now an updated cluster allocation it puts now revise cluster number three yeah beautiful and advice here cluster number five here is now reduced here to the firy paradoxon entropic principle and the dark matter and finally finally after all of this we have now the allocation that even if I have I look at it I say okay theoretically this could be a cluster allocation and I just say hey make sure that we have allocated all 20 sentences goes there python code 20 tells me those are my five clusters and those are the sentences and we have here 20 sentences and after all of this I go now and tell here my gbd4 system listen now you create an improved prompt for the exact same task like we just we did now but be as precise that next time you will find a perfect allocation of 20 random sentences to five self-defining thematic cluster without any of the iteration we went through just right now in our conversation the J gbt goes in and says okay and now Jud gbt generates based on our experience here the perfect prompt say okay these are the task I will have an initial review I will draft some cluster teams I will have a detailed analysis for the allocation I will allocate the sentences to the cluster I will then refine the cluster titles I will review it and adjust it I will run a verification and the output will be the final allocation of 20 sentences to five ref find thematic clusters and this is it so now what I do I take exactly this prompt that that GPT 4 generated for me based on our interaction up until now and now we have a working good performant prompt and you see this prompt is now really special and I say okay so the new task for you generate 20 random sentences and then I now I just copy here the new improved prompt that jet gbd4 generated and I say let's do this so first step generate 20 random sentences and now jet GPT starts goes in Python says generate 20 random sentences beautiful 20 sentences internal review draft draft cluster teams gives me five teams detailed analysis yes of course goes to the analysis 1 2 3 4 5 does a verification and this is it and just say Okay count the numbers do we have 20 it comes back and says yes we have exactly 20 sentences allocated and this is here the distribution of our five titles and if if we now have a look physical science and material Innovations relativity photosynthesis in Plants physical sciences okay nanotchnology okay biomedical research and genetic engineering crisper cost nine gut microbiome and protein folding okay astronomical and cosmology gravitational waves hble Space Telescope Dark Matter aop planets are okay environmental sciences and sustainability climate change Energy Solution renewable energy sources depletion of ozone layer Implement implications for the Earth climate Okay and computational and technological advancement very broad topic AI quantum computer machine learning techniques blockchain technology autonomous vehicles and neural network brain infrastructure so works you see now here with this perfect perfect with if this operational prompt we succeed and therefore in this simple example if you want to find a good prompt you do not need a template you do not need thep programming here just do like I do here if I have a problem look at this what we had to go through so next time when I have 20 sentences and I want that there is a cluster generated that te itic cluster by gp4 now we have a functionable prompt to do this without any programming without any problem just follow how you would talk to the computer if there's a problem tell him hey that's wrong what do you think about this and as you can see in this very short example gbd4 can self-optimize if it learned from your interaction exactly what you want from it but