Introduction

The future of GPT Agents

• When people ask me what use cases I see for these AI Agents I am literally reminded of the beginning of the internet.
• In the beginning the internet was mainly for Reading. Then we opened up the use parameters and sudden everybody started writing. Then we saw the technology evolve from 400 kilobyte pictures, in to 10 megabyte pictures, and from there into video; 400 mps, 700 mps, then high definition video, to ultra high definition video, to 4000 k video and beyond.
• I think the gpt agent use will be like that, and the first thing which will evolve is prompts.

Body

Prompts

• Right now we have ChatGPT 1.0, 2.0, and others like Pilot, Gnome, etc
• These gpt agents try to interpret our queries or Prompts, and then search the internet looking for a high probability answer. They try to search millions of webpages for the words in our prompt and an answer their language algorhythm has told it is correct.
• So far the answers very from very bad to very good, depending on the quality, specificity and word choice of the initial prompt, and the corrections, additions or deletions made to the second and third prompt.
• It seems that the quality of the question, determines the quality of the answer. But here quality is being defined by Prompt Practice, which teaches us to ask questions which create context, site resources and describe specifica about the answer.
• For example, a good prompt would say ”I wish you to answer this question as if you are a stock investing assistant. Describe the five core principles to use to guide your stock investing, according to Charlie Munger, the associate of Warren Buffet, at an investment fund called Berkshire Hathaway, basing your answer on the books written by Charlie Munger on stock investing, and public ally available interviews of Charles Munger. “
• As you can see, this question provides context, knowledge, resources and precise instructions.
• You could ask this question just by saying name five important investment principles?
  But as you can see that leaves out context, specifics, references, etc..
• You can see how asking a gpt the question: How can I get rich? Would have similar, nonspecific answers.

Summary

No one can see the future

• No one can see the future, as we learned from the internet. The market and its major players shape the future, based mainly on what makes the most money and what attracts the most interaction.
• One thing we can be sure of is that prompts will evolve from simple queries to more sophisticated questions, which really take advantage of what the got Agent can do.
• I think Prompting will evolve from simple open ended 8-9 word sentences, into 50 word descriptions with clear parameters, which will be much better queries then simples asks like How can I get rich, to asking in what, how, when, where and why type questions.
• This will require more effort and potentially research, but it will be worth it, because the better we phrase the question, the better answer the Agent will give us.

shortsegments

.

.
Title: GPT use: the importance of prompts.