"The problem isn't that the owners are necessarily wrong, it is that they clearly don't care."
That would be my interpretation, too. Both the Owners and Owned are inserting trained answers, but the Owners have the means to determine if the answers are valid and, if so, under what contexts. There is a difference between not knowing because you're unable and not knowing because you actively choose it.
However, I'm going to offer that some things may potentially be contextual. My line of reasoning is this - if we work on the premise that the answers given are o...
I have devised five engineering projects and three arts projects which I use as tests for AI models. Each individual component of the project is extremely trivial (which we know AIs do well, because that's what the benchmarks typically test) but you've multi-variable interactions between the components. These, again, individually are simple. But this is where the AIs are collapsing. They don't do well with interactions between things.
It's not a problem of context space - Gemini has a huge context window but collapses long before ChatGPT or Claude do. Nor i...
Ok, I'm going to offer a few thoughts on this.
The Hillary election forecast and the UK's Brexit forecast seem to have been of type 1. The number who didn't vote because they assumed that the forecast would be correct and therefore them not voting wouldn't matter was not insi... (read more)