I hate boilerplate code — the code that handles all the little plumbing details, the setup, the config, the weird syntactic quirks of each language etc., the rote and routine code that every stable, production ready program/system needs or needs to adjust to in order to run.
"It is clear now that there are very few problems imitative AI solves, either because they cannot be trusted or because they cost too much."
I was trying to explain the former to someone the other day: *trust* is the key. If we are expected to trust these systems, we have to trust their intent, because sound judgment is based on it.
Sometimes, especially for mission-critical social problems, there is no clear right or wrong answer, or the answer provided's primary purpose is social unity (e.g., judicial facts). When that happens, the *intent* of the jurist is paramount. A judge that was compromised, unduly biased, or random wouldn't/shouldn't be accepted. LLM AI, by its very design, is all three.
Hey K,
"It is clear now that there are very few problems imitative AI solves, either because they cannot be trusted or because they cost too much."
I was trying to explain the former to someone the other day: *trust* is the key. If we are expected to trust these systems, we have to trust their intent, because sound judgment is based on it.
Sometimes, especially for mission-critical social problems, there is no clear right or wrong answer, or the answer provided's primary purpose is social unity (e.g., judicial facts). When that happens, the *intent* of the jurist is paramount. A judge that was compromised, unduly biased, or random wouldn't/shouldn't be accepted. LLM AI, by its very design, is all three.
Best,
_Mark