For quick disclaimer, I don’t use AI to write, or make art (because I love to do these things, why would I give them up?), but I do use AI occasionally in teams, for number analysis, and to handle coding, like whipping together this website’s WordPress theme.
Skip down the page to find out how to set a values anchor to your AI. You know, so your AI does things based on what matters to you, not only on what matters to everyone else who has trained the thing to date, and all the existing materials on the web (blech, I’m not sure what most developers were actually thinking).
Large Language Models (LLMs) are set up to be extremely efficient, to find patterns, to respond to prompts, and to write and format in ways that help convey information based mostly upon what’s already out there. AI in turn produces road maps, directives, to do lists, crappy literature, bullet points galore, slide decks and generative images and video. And lots of slop.
But AI cannot necessarily understand why we do what we do, or how we choose to do we do, or why things were done in that order.
And this is incredibly important– you know, narrative.
The why and how, our values and order order point to our human “becoming goals.” And arguably the unique constellation of forces: the what, why, in what order and what circumstance, is completely unique for every individual and also, an aspirational pointer toward our best selves.
Unless you take time to write something like an annotated Standard Operating Procedure (SOP) of your life, with emotional cues, irrational movements, serendipity, and your desires and fears, AI cannot possibly give you satisfying results that are correctly aligned to YOU. But, AI may be able to do a decent, practical, support job if it knows your mission, values, and goals.
Also, at the moment, AI can’t truly wonder, what would help our planet and it’s life forms thrive on the whole? But maybe you do?
I hope so, because this aggregate of content and coding parameters called AI likely needs people like you to tilt the training materials for good.
These days AI appears to be producing substance and slop because of assumptions that humans always want more speed/efficiency/time/options/money/directives/formatting and content. But, humans I think, actually want more simplicity.
So, are the most important aspects of you reflected in the output you get with this tool? If not, now is the time to at least attempt to get clear about what’s important to you and share it there.
Not only will this process help you understand yourself, it will also train the model on some higher quality aspirations, which it surely needs.
Here’s how to start prompting around your finest aspirational self in the form of parlance around “mission” “values” “purpose” and maybe even your unique “North Stars.”1
Once you identify your Top Three Values, for instance, you can “anchor” your AI agent with this core material. This trains your AI to keep these positive values front and center, and to store them in memory. Even if you guess what your values are, based on the best things in life you’ve made, that’s better than avoiding this “anchoring” task.
In Claude, for instance, you can turn on the profile>>setting for “memory” and create a Project called “Anchor Values.” There you write your key values, aspirational goals, authentic desires for self and community, and ask it to keep this framework of values & goals at the fore for all output thereafter. You can even direct it to question tasks that may fall outside of these important-t0-you parameters. That’s useful, don’t you think?
If you are on another AI, ask it where is best to enter a values-based set of directives to serve as a foundational directive and framework for all other chats.
If you’d like more around this kind of material, drop me a line.
There are many, many specific prompts that I could share that I’ve either developed or copped from other folks. But honestly, who wants the smartest sausage out of a sausage machine? It’s still made of a pile of parts that you may not eat otherwise. Time to up the input, or at least make it relevant to you.
Garbage in, garbage out. And for you, I wish for quality in, quality out. Also, hoping the same for AI in general, since it’s likely not going anywhere soon.
FOOTNOTES & SUB-SUBTEXT ;)
- And please don’t ask AI to give these things to you. These must come from your heart and soul. ↩︎