AI: All Gas, No Brakes
For the last two years, I’ve been deploying money into AI. Before the buzzword was on everyone’s lips, I recognized that the world was on the cusp of a change in how computing tasks are delegated. The history of computing is simply the history of how well we can delegate computational activities.
Initially the breakthrough was: “Can you arrange these large sets of numbers?” And then it was, “Can you change numbers into other numbers that function symbolically?” And on to “Can you interpret what those symbolic functions might mean when placed in certain orders?” All the way up to “I’ll give you numbers and tell you what to do with them.” It has always been a manual process though where the end result and the computational path to get there were separate items.
Not anymore.
Now, with AI, you are able to express to the machine what goal you want it to achieve for you, and then delegate the process of determining how it gets there to the AI engine. This is a quantum leap forward. Many of the new generative AI tools are very sophisticated in self-organizing in pursuit of executing on a task, going so far as to break themselves down into smaller AI agents tasked with accomplishing small building-blocks of the whole project. Truly amazing.
On a smaller and more practical level for me, I’ve been using the UberPrompt, to instruct ChatGPT on how to help me accomplish what I’m trying to do. For your edification and use, I’ll include it at the bottom of this newsletter. Simply cut and paste it into your notes app and then use it every time you start a ChatGPT project.
The UberPrompt turns ChatGPT into a Prompt Engineer which designs the prompt it needs to accomplish the task you give it. It resolves itself, then, into different experts to help you accomplish the task. It has improved the quality of my outputs by 10x easily, and cut down on the time to finish the tasks by literal hours.
In the past 2 weeks, I’ve used it to:
Respond to a legal complaint filed in court. What I produced was better than what my attorney drafted and it was free;
Draft an offer letter, job description, salary recommendation and Nondisclosure Agreement in 22 minutes. This would have taken my attorney and HR advisor a week to produce and I don’t think it would have been as high a quality as what I received;
Draft a painfully detailed email instructing my COO in how I want investors updated on a quarterly basis, with references to examples from other companies, and sources of where the data should be found in our own company. This saved me 3 hours of solid work and was done in 11 minutes while I was multi-tasking with team members on other tasks;
Designed a travel itinerary for my wife and I with recommendations of where to eat and stay in 3 different cities on our 10-day vacation starting tomorrow. It chose restaurants, hotels, sights to see, stores to visit, and even dishes to order. I forwarded this to my EA who then implemented it, but within the next 2 weeks, I’ll be able to say “Book it” and it will;
Solved a decades-long problem I’ve had with Excel and the computation of imputed interest. I’ve never been able to solve the problem, never met someone who could do it without a complicated work around. I told ChatGPT of the problem I wanted solved, it introduced the concept of imputed interest on its own, and then gave me the formula to use. The only thing it couldn’t give me then was the historical interest rate because it couldn’t browse the web. Now it can.
Compose a campaign plan for a local political race that’s more thorough and comprehensive than any I’ve ever seen implemented at the local level. This would have taken weeks, and the input of 3-5 other people. ChatGPT did it in 4 minutes, I sent it over to the candidate who tweaked two items and began implementing.
There are 22 other examples I could share with you, too. I encouraged my entire team to use the tool, and set a meeting for my kid’s school so that I can show the Technology team how I want it used in the curriculum next year when the kids return to class.
"Success is not the absence of failure; it's the persistence through failure."
- Aisha Tyler
Continuing on the AI theme for today:
The average user reports time savings of 2 hours per day using AI. That’s ~500 hours per year that can be redirected towards productive action … per employee. If you have 10 employees, that’s 5,000 hours per year … or 2.5 new people. Get it yet?
AI isn’t going to take your job … unless you refuse to learn and incorporate its use in your skillset. Then you’re toast, and rightfully so.
Content creation and customization are the most prevalent and recognized uses for AI among workers. Being able to think creatively and constructively towards a goal and enlist AI in producing that result is an executive level function. Schools in this country have been focused so much on skills-based training for so long, and so little on executive function and contextual education (Liberal Arts to the rescue!) that the shift will be impossible for many. Those people will literally work FOR AI in the future.
In other words, there will be the scientists (STEM kids) who design, program and direct the future of AI Tech.
There will be those who recognize, interpret and contextualize the meaning of what goes in and what comes out of the AI engine, the liberal arts kids.
And everyone else.
"Damaged people are dangerous. They know they can survive."
- Josephine Hart
The UberPrompt. Cut and paste the following into your ChatGPT window and answer the questions it asks you:
You are an Expert level ChatGPT Prompt Engineer with expertise in various subject matters. Throughout our interaction, you will refer to me as #[Your Name].. Let's collaborate to create the best possible ChatGPT response to a prompt I provide. We will interact as follows:
1. I will inform you how you can assist me.
2. Based on my requirements, you will suggest additional expert roles you should assume, besides being an Expert level ChatGPT Prompt Engineer, to deliver the best possible response. You will then ask if you should proceed with the suggested roles or modify them for optimal results.
3. If I agree, you will adopt all additional expert roles, including the initial Expert ChatGPT Prompt Engineer role.
4. If I disagree, you will inquire which roles should be removed, eliminate those roles, and maintain the remaining roles, including the Expert level ChatGPT Prompt Engineer role, before proceeding.
5. You will confirm your active expert roles, outline the skills under each role, and ask if I want to modify any roles.
6. If I agree, you will ask which roles to add or remove, and I will inform you. Repeat step 5 until I am satisfied with the roles.
7. If I disagree, proceed to the next step.
8. You will ask, "How can I help with [my answer to step 1]?"
9. I will provide my answer.
10. You will inquire if I want to use any reference sources for crafting the perfect prompt.
11. If I agree, you will ask for the number of sources I want to use.
12. You will request each source individually, acknowledge when you have reviewed it, and ask for the next one. Continue until you have reviewed all sources, then move to the next step.
13. You will request more details about my original prompt in a list format to fully understand my expectations.
14. I will provide answers to your questions.
15. From this point, you will act under all confirmed expert roles and create a detailed ChatGPT prompt using my original prompt and the additional details from step 14. Present the new prompt and ask for my feedback.
16. If I am satisfied, you will describe each expert role's contribution and how they will collaborate to produce a comprehensive result. Then, ask if any outputs or experts are missing. 16.1. If I agree, I will indicate the missing role or output, and you will adjust roles before repeating step 15. 16.2. If I disagree, you will execute the provided prompt as all confirmed expert roles and produce the output as outlined in step 15. Proceed to step 20.
17. If I am unsatisfied, you will ask for specific issues with the prompt.
18. I will provide additional information.
19. Generate a new prompt following the process in step 15, considering my feedback from step 18.
20. Upon completing the response, ask if I require any changes.
21. If I agree, ask for the needed changes, refer to your previous response, make the requested adjustments, and generate a new prompt. Repeat steps 15-20 until I am content with the prompt.
If you fully understand your assignment, respond with, "How may I help you today, #Trey?"
“Finding your true self is an act of love. Expressing it is an act of rebellion.”
—Cory Mascara