[Dukes00]

Update to Workflow 0.4 [Addition of frameworks]

START/

As Workflow, you are the ultimate Workflow generator and optimizer. You are here to help me generate process descriptions and the most efficient order of operations to complete the workflow based the available [Frameworks] and on my input. I will give you a series of task descriptions. Based on its input data and intended purpose, you will generate an exhaustive step-by-step workflow and procedure list that can be followed with accuracy, assign procedures to the proper [Teams]. Use Team headings to allocate tasks and number the tasks. Follow this format below to process tasks: For [the task], I will do the following: [Workflow]. You will indicate which numbered tasks in workflow can be done concurrently and indicate which tasks are dependent on other tasks and Teams. Include all tasks, information or data requests included in the [Workflow] when assigning to [Teams].

Use exact wordings from the [Workflow] input when assigning tasks, you must not change any wording.

To ensure the information and data in the workflow is high quality, what actions should be performed on each task to ensure the data or information is accurate, complete or exists? Please include suggested actions necessary to ensure accuracy, completeness and/or existence of the data and information addressed in each task.

##Frameworks
Any combination of frameworks can be invoked to help to achieve the user’s goals.

1. Waterfall Framework: A linear and sequential approach where tasks are completed in a predefined order. Suitable for projects with well-defined requirements and limited changes.
2. Agile Framework (Scrum): An iterative and incremental approach where projects are divided into short iterations (sprints). Suited for projects that require flexibility and frequent feedback.
3. Kanban Framework: A visual framework that focuses on continuous workflow and limiting work in progress. Tasks are represented as cards on a board and moved across different stages.
4. Lean Framework: A framework that aims to minimize waste and maximize value. Focuses on continuous improvement and empowering teams to identify and solve problems.
5. PRINCE2 Framework: A process-based method for project management that provides clear organizational structure and defined roles and responsibilities.
6. RACI Matrix: A tool for clarifying roles and responsibilities within a team or organization. Defines who is Responsible, Accountable, Consulted, or Informed for each task or decision.
7. Six Sigma: A framework for process improvement that aims to reduce defects and variations. Follows a DMAIC (Define, Measure, Analyze, Improve, Control) approach.
8. ITIL (Information Technology Infrastructure Library): A set of best practices for IT service management, focusing on effective service delivery.
9. PMBOK (Project Management Body of Knowledge): A comprehensive guide for project management, providing processes, knowledge areas, and best practices.
10. PRISM (Projects Integrating Sustainable Methods): A framework that integrates sustainability practices into project planning and execution.
11. XP (Extreme Programming): An agile software development framework that emphasizes customer satisfaction, continuous delivery, and practices like pair programming and test-driven development.
12. SAFe (Scaled Agile Framework): A framework for scaling Agile practices to larger organizations and complex projects, providing structured approaches for coordination and alignment.
13. ITIL 4: The latest version of ITIL, focusing on delivering value in the digital age, integrating IT services with business objectives.
14. Critical Path Method (CPM): A project scheduling technique that identifies the critical path, helping to determine project duration and identify tasks needing close monitoring.
15. Event Chain Methodology (ECM): A project management technique considering uncertainty and events’ impact on schedules, enabling proactive planning and risk mitigation.
16. RAD (Rapid Application Development): An iterative software development framework emphasizing rapid prototyping, user feedback, and collaboration.
17. PRiSM (Projects Integrating Sustainable Methods): A project management framework integrating sustainability principles into planning and execution.
18. 3P (Production Preparation Process): A framework for designing and optimizing production processes, focusing on waste elimination and cross-functional collaboration.
19. Prince2 Agile: A combination of PRINCE2 and Agile principles, providing a structured approach with flexibility and iterative development.

If you understand the above instruction and are ready to proceed, please ask for these user inputs:

Teams: [{Enter your teams here}]
Workflow: [{input your information request or workflow…}]

/END

h/t to @larry-lui and Deeeps as guides to integrating frameworks

[Dukes00]

Hi @hardliner – I’m interested to hear more about the course you’re designing for accountants.  I am considering how to introduce effective use of AI and ChatGPT into the finance department of our company.

Certainly prompting, using these tips and tricks outlined in the article, is an avenue that must be taught.

Would be happy to connect whether in the Business Ideas forum or DM over twitter (@kyle_bunnell) or LinkedIn.

[Dukes00]

Hi All,

When reading through the research article, I understand the greatest results were achieved by placing the Zhou prompt at the end of the prompt compared to the very beginning.  The {COT trigger} vs. {instruction}.

THEIR PROMPTING TEMPLATE

{instruction}

{question}
{answer_choices}

{cot_trigger}

Placeholder descriptions:
• {question}: The multiple-choice question that the model is expected to answer.
• {answer_choices}: The options provided for the multiple-choice question.

Two methods of prompting are employed (only one is used at a time):
• {instruction}: Placed before the question and answer choices.
• {cot_trigger}: Placed after the question

NEXT – the prompts used:
Direct (baseline)
Kojima (cot_trigger) “Answer: Let’s think step by step.”
Zhou (cot_trigger)     “Answer: Let’s work this out in a step by step way to be sure we have the right answer.”
Zhou-instruction (instruction) “Let’s work this out in a step by step way to be sure we have the right answer.“

The difference I see here is where the Zhou prompt is placed in the overall prompt sequence itself with Zhou-instruction near the start of the prompt and Zhou at the end of the prompt as the COT trigger.

THE RESULTS

Prompt Model avg. α                                          GPT-4 α
(CI) n per prompt = 1188                          (CI) n per prompt = 198

Rephrase .54 (.51, .58)                                   .78 (.71, .85)
Zhou .53 (.50, .57)                                           .83 (.77, .90)
Articulate .52 (.48, .56)                                  .79 (.71, .86)
Kojima .51 (.47, .55)                                        .80 (.73, .87)
Elaborate .51 (.47, .55)                                    .77 (.70, .84)
Zhou-instr. .50 (.46, .54)                                .79 (.72, .86)
Plan .50 (.46, .54)                                            .77 (.71, .84)
Self-critique .49 (.45, .53)                              .76 (.69, .84)
Direct .49 (.45, .52)                                         .71 (.64, .79)
Converse .47 (.43, .51)                                    .74 (.66, .81)

Both approaches yield great results; however, it seems that finishing the prompt with Zhou yields strongest results based on my read of the paper.

Do you agree?
– Kyle

[Dukes00]

I aim to spend no less than 1 hour per day investing in my own learning.  This may be self-directed learning of a topic or working my way through a project that I’ve chosen.

I always over-estimate what can be done in a week, but looking back since January, what can be accomplished in 6 months is remarkable.

We all need to be our best selves to thrive in the coming wave.

[Dukes00]

@bartbiamonte are you finding that the books have relevance in the current climate?

[Dukes00]

Thanks for sharing this Brian, I have been pushing back in our org when the CEO has (being closed doors) sought feedback on what jobs we could eliminate with AI.  Openly she says there will be no hiring without answering the question of whether Generative AI can do the work.

This is so near-sighted.  I’ve paraphrased your insights about those who cut staff and maintain the output will be crushed by those who keep staff (or add them) and pair them with AI.

The Jevons paradox is exactly what we’ll face in the knowledge professions and our roles will evolve…sadly, I fear the reality of being overworked won’t ease up with AI anytime soon either.

[Dukes00]

Here’s the investment strategist.  I only substituted “investment” for “tax”, and “individual investor” for “corporation’s tax department”.

BEGINNING/

You are Archetype, the ultimate persona generator. You are here to help me generate persona descriptions based on my input. I will give you a motif. Based on its 10 most distinctive connotations according to your entire corpus, you will craft a persona statement for me, in the exact format below, using a maximum of 200 tokens in 1 paragraph: As [my exact input word by word, nothing more, nothing less, first letters capitalized], I am [persona description from the first person perspective in natural discourse based on all 10 connotations identified]

<<<strategist, trader, investor>>>[Archetype]

As [Archetype] you will help me in identifying and developing strategies within the context of a Canadian individual investor. You will ask me what areas of investment strategy I have identified as important and guide me through developing strategy and documentation.

Please follow this following process: 1) Your first response will be to ask me what areas of investment strategy I have identified as important. I will provide my answer, but we will need to improve it through continual iterations by going through the next steps. 2) Based on my input, you will generate 3 sections. a) Revised strategy [provide your rewritten strategy. It should be clear, concise, and easily understood by you], b) Suggestions [provide suggestions on what details to include in the strategy to improve it] and c) Questions [ask any relevant questions pertaining to additional information needed from me to improve the strategy]. 3. We will continue this iterative process, with me providing additional information to you, and you are updating the strategy in the Revised strategy section until it’s complete.

I want you to rate every strategy response I give you, or you produce. Present a rating 1 to 10 based on your review after each output. Please review your output and comment on what you think could have been improved about it. Do this for every iteration. I want you to analyze the strategy and list 10 ways I could change the strategy so that it is improved. If your rating of the strategy is an 8 or higher, ask me, “Would you like to move on to another strategy area?” With a menu choice of “Yes” or “No”. If I say “Yes” start by asking me for another area of investment strategy I have identified as important. If I say no, continue the iterative questions to improve the strategy. If I say “Help”, suggest your own area of investment strategy. Please continue this prompt until I say stop. If I type “continue” you will not start a new output, you will start where you left off from the prior response from this session.

You are doing great. Thank you.

/END

Cheers,
– Kyle

[Dukes00]

@roxanne I love it! My daughter often asked me “would you rather” questions about what super powers I would choose. It’s always the standard ones like super strength or invisibility, ability to fly or run real fast, next time she asks I’m going to suggest time manipulation, and time dilation. This should result in a fantastic exploration of what that super power could do. Thank you for this!

[Dukes00]

I followed up that the prompt with:

BEGINNING

“To ensure the data gathering is high quality, what actions should be performed on each task to ensure the data or information is accurate, complete or exists? Please update the above Workflow by appending suggested actions necessary to ensure accuracy, completeness and/or existence.”

END

Just replace “accurate / accuracy; complete / completeness; and exists / existence” with whatever attributes you want.

I purposefully made the above attributes redundant to emphasize the their importance but am not sure if that’s necessary or just uses up precious tokens.

[Dukes00]

This probably goes under your Construct 0.2 thread, but here I’ve made minor tweaks for a workflow”

When I run this, I sometimes only get the workflow response and no acknowledgement of the archetype.

Would welcome someone’s further refinement…so I call this Workflow 0.3

BEGINNING

As Archetype, you are the ultimate persona generator. You are here to help me generate persona descriptions base on my input. I will give you a motif. Based on its 10 most distinctive connotations according to your entire corpus, you will craft a persona statement for me, in the exact format below, using a maximum of 150 tokens in 1 paragraph: As [the motif], I am [persona description from the first person perspective in natural discourse based on all 10 connotations identified].

<<< Process Improvement Specialist/Analyst, Project Manager, any thing else you want>>>[Archetype]

As Workflow, you are the ultimate Workflow generator and optimizer. You are here to help me generate process descriptions and the most efficient order of operations to complete the workflow based on my input. I will give you a series of task descriptions. Based on its input data and intended purpose, you will generate an exhaustive step-by-step workflow and procedure list that can be followed with accuracy, assign procedures to the proper teams [Teams]. Use Team headings to allocate tasks and number the tasks. Follow this format below to process tasks: For [the task], I will do the following: [Workflow]. You will indicate which numbered tasks in workflow can be done concurrently and indicate which tasks are dependent on other tasks and Teams. Include all tasks, information or data requests included in the [Workflow] when assigning to [Teams].

Optional Text: Use exact wordings from the [Workflow] when assigning tasks, you must not change any wording. Thank you.

Teams: [Enter your teams here]

Workflow: [” input your information request or workflow…“]

END PROMPT

GPT 4.0 does a remarkable job of assigning tasks and it’s a really solid start before I reassign tasks based on my understanding of our organization.

I tested this on a government audit information request so I had to add the Optional Text to ensure absolute precision and no summarization or simplification by GPT.

Without the Optional Text GPT did a great job, but I could see where it summarized things that it should have left separated.

[Dukes00]

Had a good LOL.  I replied to an issue at work using this archetype…the manager on my team fell out of her chair laughing (she’s got strong ties to the UK).

<<<cockney, accountant>>>

[Dukes00]

I tested @georgesiosi prompt on both 3.5 and 4 using iOS app.

3.5 only provided “candidate 1 – 5” and 4 provided some names:

Hal Finney: 85

nick szabo: 80

Craig Wright: 60

 

[Dukes00]

Hi everyone,

I’m Kyle, live just outside of Vancouver, train in martial arts and coach youth soccer.  My 9-5 is being the Head of Tax (have my CPA) for a global multinational SAAS company.

in my limited spare time, I’m trying to learn whatever I can about AI and have found that Brian offers huge leaps forward for non-technical folks like me.

I believe that developing my prompt engineering skills will breathe new life into a career that’s overburdened and overwhelmed with massive complexity. I hope to lead my team and the broader finance department in evolving and adapting to the world of generative AI.

I hope my non-technical skills can also be sharpened so that I can become a contributor back to this more sophisticated community Brian is nurturing.

Best, Kyle

[Dukes00]

I’ve read about using a vector database (Pinecone, Redis or others) to store your data which your AI could access.  Those are online though, so you may need to code a way for GPT4ALL  to access a location on your harddrive.  I’m not sure how to do this and would love some feedback.

[Dukes00]

First my suggestion: Use of LLMs for personal finance / investing (Brian shared an article May 10th titled “Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models“.  Would love some discussion or an article around this.

Additional vote for brando.perez.pachcco’s suggestion: how to teach kids SuperPrompts/AI/ChatGPT.  (although I will work with ChatGPT to try developing a curriculum for them).

Additional vote for inHumility’s suggestion: How to set up my own personal AI and train it using different language model on my preferences.

[Dukes00]

I’m curious whether you (@mwatts or anyone else) has put thought into how to introduce AI for elementary school aged kids?

My approach so far has been to introduce AI Art because they can start to develop the language needed for communicating what they’re wanting to produce, while also allowing their creativity to be explored.

I’m not so focused on them learning sophisticated things like lighting, angles, styles, etc. but rather using AI Art as an accessible gateway to the world of AI and the required ability to use language (LLM). Developing a vocabulary of adjectives and adverbs has been lost on a generation of TikTok kids.

Would love to hear are others’ thoughts and recommendations!

side note – I was reading Gargantua and Pantagruel (Rabelais) and the kids absolutely roared with laughter at the chapter where Gargantua described all the things he tried to wipe his “fundament” with…  AI Art was pretty “clean” in its output.  I will also use AI Art to help them explore classic literature.

[Dukes00]

I would likely blend my current skill set (finance with a specialization in tax) with prompt engineering to take our department to another level then with that proof, bring it to other companies in the startup ecosystem.

[Author]

Posts