8:15a hydrate+caf
9:30 protein cocoa, oo, vitamin+
:: at least 10:30 before heading to gym, but more likely post 1pm bc that's even lighter usage time
gym
grocery
* * *
Much of what's being marketed as "AI" in the past few years would not have met the standards for the term as recently as 2020. Before the boom in LLM-backed chatbots and "generative AI" toys, "AI" was a designation reserved for research into useful approximations of human capacity to integrate new information, adapt, and apply reasoning.
What is "reasoning"?
- Logic, e.g. "stress/trauma can underlie vomiting, but probably not sinus congestion or swollen feet"
- Evidence, e.g. proving diagnostic data preparatory to rendering a tentative medical diagnosis
- Context, e.g. "the patient lives in crowded, high-traffic university campus conditions"
So-called AI is not great at this "evidence" part: a doctor can poke a patient's foot firmly with a pencil to measure reflexes, and the seemingly simple, common human experience of viewing and interpreting the patient's bodily motion (or lack thereof) as "responsive to touch" involves staggeringly intricate brain processes. It takes a lot of image-analysis technical expertise and compute resources to enable machines to do what human vision and cognition do in moments, even in narrowly contrived artificial contexts like solving a CAPTCHA.
This is also why we don't really have "self-driving cars" yet, and why Alphabet's prove-you're-a-human CAPTCHAs are completely obsessed with roadway navigation phenomena: traffic lights, cars, buses, bridges, crosswalks, bicycles, motorcycles, stairs. It is 100% not coincidence that their CAPTCHA elements both draw from, and refine next-level leveraging of, the massive Maps data source.