Feb 14
My ChatGPT / Agile Project Management Diary
What is this about?
Since early February 2023, I’ve been exploring the extreme highs and lows of the ChatGPT interface into the GPT-3.x AI Model. More recently, I’ve begun using the GPT-4 model for some tests (it’s still heavily rate limited).
I’ve been checking out multiple applications with increasing interest and refinement. One (obvious) application is in project management of the more agile variety.
Although my GPT journey started recently, the prospect of solid benefits in planning and executing AI-supported projects through this interface is tantalising.
But to make this work, I need to put some structure around the investigation, develop a method, and test out some (many) hypotheses. So I plan to share these with you as I go and get your input, feedback, suggestions and questions.
These posts might be raw and appear random: I think and do this as I go. I’m also experimenting with how to represent this in a blog post.
For now, I’ll structure it as the raw question and answer.
This introductory yarn is now the master index page for these diary entries — see below — and I’m using ChatGPT to help create the summaries — see “Spoiler Alert” below.
Diary Master Index
Spoiler alert: I’m using ChatGPT to generate 50-word summaries of each yarn after publication. As I mention in Part #6, there are no cases where I can copy/pasta the ChatGPT output into the page and publish. However, ChatGPT gives me a good starting point. Edits range from minor tweaks to major rewriting.
This yarn seems way off-topic but illustrates an important aspect of using AI models. I tried to use the GPT-4 language model to help me solve the word game Wordle, but it provided wildly incorrect suggestions. I tried a second time with more structured prompts and establishing ground rules with GPT-4, but the results were worse. This article emphasizes understanding the strengths and limitations of language models and how to use them effectively, raising questions about their practicality for simple tasks.
This post explores Prompt Engineering for ChatGPT from the perspective that writing prompts is writing code and can benefit from incorporating Agile and software development principles. I use Peter Naur’s concept of coding as “theory building” to argue that developing a strong “theory” or “mental model” of the problem is essential for successful prompt development. It also emphasizes the importance of separating the real world, models of the real world, and the systems and solutions that implement those models. Ultimately, the post highlights the value of the “theory building” approach to allow easy transitioning to new AI tools in the future.
This yarn explores the short-sightedness of canned prompts in ChatGPT, warning against copying and pasting pre-written prompts without understanding how they work. I emphasise the importance of Prompt Engineering and the need to learn about the tool to avoid the “Zone of Suck”. I compare ChatGPT’s prompt types to generic NLP prompt types. I also provide a link to two tables generated by ChatGPT (from my prompts) that exhaustively list the 60 ChatGPT prompt types and 38 generic Natural Language Processing prompt types.
ChatGPT / Agile #9: Prompt Engineering | by Adam On Projects | Mar, 2023 | Medium
This article looks at “Prompt Engineering” in the context of natural language processing systems, discussing the importance of designing and optimizing prompts to guide the output of language models like ChatGPT. The article provides nine online references to books and stories about “Prompt Engineering”, emphasizing the need for context and nuance when interacting with AI models. Improving prompt engineering skills can lead to more tailored and accurate responses from ChatGPT.
This yarn focuses on the limitations of AI language models such as ChatGPT, which cannot think in any way related to human thought. Machine learning programs operate through statistical pattern recognition. While they have wide-ranging and powerful capabilities, they lack the critical capacity to explain causation and distinguish between the possible and impossible. The article emphasises the importance of expert verification and control in using ChatGPT. It can produce skilled and flawed content that cannot be relied on for all decision-making.
This yarn describes my argument with ChatGPT over creating a fictional Scrum team with diverse demographic attributes. ChatGPT initially objects to including such attributes due to potential discrimination and stereotyping concerns but ultimately generates a list of team members with the requested characteristics. The dialogue highlights the complexities and ethical considerations of using AI language models and the importance of being clear and specific in requests.
Part #6 reflects on the capabilities and limitations of ChatGPT in agile project management. I discuss using ChatGPT in a live application or use case and share their experience with ChatGPT’s outputs. The article concludes by highlighting ChatGPT’s strengths and how it can provide the scaffolding to “bootstrap yourself to badass” if used actively.
Part #5 extensively explores how ChatGPT can help with Agile User Stories. The article provides many practical examples and offers insights into how ChatGPT can help generate, analyse and manage user stories in Agile projects. I also introduce a new technique for managing text in a ChatGPT session and identify at least one limitation.
Part #4 digs into the analogy of the Scrum agile framework to the game of Rugby, where the term originated. On one level a light-hearted proposition that the originators used the wrong Rugby formation to name the framework. But more importantly, this explores ChatGPT handles very subtle nuances of language and words that have the same form but vastly different meanings. It also illustrates how ChatGPT and users can fine-tune their dialogue for better results.
Part #3 explores the agile knowledge areas and how they can be leveraged in project management. The yarn illustrates how ChatGPT can be a natural language interface to a fact base.
My ChatGTP / Agile Project Management Diary — Part #2 | by Adam On Projects | Feb, 2023 | Medium
Part #2 explores the capabilities of the ChatGPT interface and its potential use in agile project management. The yarn outlines the approach I’ll be taking to investigate ChatGPT and its capabilities to enhance Agile project management workflows.
My ChatGTP / Agile Project Management Diary — Part #1 | by Adam On Projects | Feb, 2023 | Medium
Part #1 (this yarn) introduces a new series exploring the use of ChatGPT in agile project management. I discuss the potential benefits of using ChatGPT to plan and execute AI-supported projects, while acknowledging the need for structure and testing of hypotheses to balance out the breathless and shouty hysteria over ChatGPT and its potential impacts. The series will cover various topics, presented in a question and answer format, with opportunities for feedback and discussion.
