Here’s an overview of the prompt I’m currently using to play D&D with ChatGPT:

In my mind, I’ve got as new talk that evaluates several of the AIs out there based on how well they can play Dungeons and Dragons. After using ChatGPT for about a year, I have first hand experience that makes all the hype - good and negative - seems a lot less amazing. ChatGPT is practical, it’s quick, fun, and even creative. But, it’s not perfect at being a D&D dungeon master right out of the box. It takes a lot of work, adapting, and work on the player’s part. This feels like a good test case to learn and experiment with what exactly all this AI stuff can do in all domains, business or fun.

While it’s a game, Dungeons and Dragons has extensive rules, almost endless lore and commentary on the Internet, and an open ended nature that means you can do anything. To work well, it requires strict adherence to a lengthy set of rules, but a lot of judgement in how they’re applied and interpreted. And, at the same time success at Dungeons and Dragons requires a lot imagination, creativity, and, well, making shit up.

That’s just like business strategy, marketing, product development, maybe therapy, law, and most things that are high-value activities. I mean, what do us humans do except try to predict what is going to happen next and predict what should happen next, or, predict what we want to happen next?

Put another way, I think if I can get ChatGPT to play a dungeon master well, I can probably get it to do other human-creative things well.

Furthermore, there’s a lot of these generative AI things out there. How would you rate them? Usually you take the same problem/task, come up with some criteria for grading the thing’s performance, and then run that test across all the options. That’s the second part that I want to do: come up with a way of rating and judging the performance of each AI setup. In this case, when it comes to being a dungeon master.

If I come up with a few criteria and test cases (like, an adventure to play on each), I can evaluate ChatGPT versus Bard versus a Tanzu OSS stack versus Bing/CoPilot versus Claude (whenever it’s available in the EU) versus Watsonx, etc. & what have you.

We’ll see! This could also just be an elaborate excuse to play D&D instead of organizing my desk drawers.

The, uh, “analysis” of my prompt above is based on a session I did live today which you can also watch if you’re into seeing it all happen step by step. It ends right as combat starts with some goblins, which I’ll have to pick-up next time. In that livestream recording, you can see the tool-chain and technique I use for solo playing with ChatGPT, along with some commentary on the project.

If you want to check out the prompt more, it’s in the description of the video.

My Content

• Write the letters - Software Defined Talk #452: This week, we examine the balancing act CEOs face between maintaining operations and pursuing growth, the IRS's attempt to automate tax filing, and defining success in thought leadership.

• How to get over your fear of public speaking.

• For conference presentations, use your own visual style.

Relevant to your interests

• Return to Office mandates boost company profits? Nope - More studies.

• To CEOs and VCs Suggesting Tech-Staffing Cuts - He’s not in favor of the app architectures of the past 15 years. Nor kubernetes, or Scrum.

• Can Enterprise DevOps Ever Measure Up? - A check-in on developer productivity and DevOps metrics.

• Gartner Survey Finds Sales Analytics Has Less Influence on Sales Performance Than What Leadership Expected - “Eighty-four percent of sales leaders agreed that sales analytics has had less influence on sales performance than what leadership expected.”

• GitHub Copilot Research Finds “Downward Pressure on Code Quality” - The AI generates too much code and does not favor code re-use.

• AI Design Patterns - As it says! This is from a VCs perspective, collecting together the patterns they see.

Wastebook

• “He was getting his team to paint a fence.” “About ten percent of the drowned men had their flies open.” Here.

• “Repatriating the partner margins.” Here.

• “I wonder how long it would take to reach product-market-fit if I start with a double-sided spoon…” Derek Sivers.

• “I just replied to an email from 2018.” Current year: 2024. Here.

• I don't understand very much but I know a lot of things.

• “I recall hearing a presentation by a GE senior executive at a conference I attended almost ten years ago. Accompanied by a surfeit of PowerPoint, the executive spoke about how GE was reinventing itself as an information technology company. Yet the talk, with its generalities, platitudes and bureaucratic and taxonomic framing, sounded like the corporate America of a bygone age. I walked away from the session convinced that the company had no future. The successful tech people were … crazier than that, more ambitious and more focused on what their companies can and cannot do well. And they don’t use PowerPoint.” Here.

• “I’m a shower idiot!” Tyler Cowen.

• “Dungeon Crawler Carl.”

Conferences, Events, etc.

Talks I’m giving, places I’ll be, and other plans.

DevOpsDays LA, March 15th, “We Fear Change” talk; KubeCon EU, March 19th. Executive dinner in Dallas, March 31st.

Logoff

Weird.