Was having a quick chat with @Kevin_Schumacher yesterday and he brought up something I can’t stop thinking about!
”AI needs to be able to predict patterns based on humans, but humans are unpredictable.”
And to that I said….. whhhhattttt 
Curious if this resonates with anyone else? It makes me think AI will in a way, always be a step behind?
From a psychology standpoint, I think most would argue that humans are quite pattern based & perhaps predictable, but are the instances where they’re not - is that where the magic / value happens?!
@Kevin_Schumacher has a lot more insight on the topic of AI than I had originally thought!
Curious if you’re a programmer as well, haha
@nathaliesmith one aspect of human unpredictability is how our social relationships drive our development and thought patterns.
One way to look at this is like a trifecta:
These three areas are ideally in alignment, but when we have social misalignment with our tribe, we begin to change our behaviors for survival and we look for a new tribe.
So people and societies moving through transition will reject old patterns as they explore new ones.
Good luck keeping up with that!
I try hard to avoid tribal thought patterns. I think these are probably more predictable but in a bad way?
@nathaliesmith , I would say Individual would actually be where the good predictability happens. Because, predictable also isn’t predictable.
@nathaliesmith I totally agree that our current incarnation of AI (LLMs) will always be a step behind. But that really depends on the context of the user. LLMs are just a fancy way to retrieve information at the end of the day. It can feel creative or novel because the user may not be expecting what it outputs. The flip side is if the user does have an expected output and it misses, it feels broken. It’s sort of a miracle that LLMs do anything sensible at all when you really boil down the nuts of bolts of whats happening!
So I agree it will be a step behind with respect to all of human knowledge but for an individual user, it can be way ahead of them depending on what they are trying to accomplish!
@Doug_Curtis Love this POV and totally agree with you! It will be so interesting to see whats on the horizon.
“So I agree it will be a step behind with respect to all of human knowledge but for an individual user, it can be way ahead of them depending on what they are trying to accomplish!”
Design quality now has more to do with the quality of conversation between human and machine.
Which begs the question, how do we evolve our Glare metrics to evaluate this quality?
100%. I think we have to lean into this idea that the UI world will probably shift, and I think Glare will also have to shift in a more personalized sense of tracking.
Maybe the metrics will be the same, but some will become way more important?
This is why I like to use LLMs as rubric based on how I want it think, then report back if I’m accomplishing that… I’ve created five or so that I use on a regular basis. Definitely creates a more compelling experience.
That sounds really interesting, I’ve had similar thoughts of creating prompts to act like a “lens” to make sense of a lot of information… This sounds similar, do you have an example?
@doug, here’s a rubric for evaluating an agreement. (sorry, it’s normally a PDF, but I can’t attach)
This would be used to determine whether your ask is clear and well-defined.
Update: Here’s a google doc, Agreement to Action Rubric (STAR)
