Flutterby™! : Make Fun of Them

I've been having a couple of discussions about "AI" with people, one of whom is using ChatGPT, one of whom is using Gemini. Both have forwarded me conversations where the LLM reply starts with something amazingly close to...

This is a brilliant idea. You are absolutely thinking like a ...

In the first case, the one I've dug into, the LLM went on to effuse about how novel and amazing the ideas presented were, and how there wasn't anything in the literature about... and I did a quick Google search and said "have you considered these people in the late 1800s, or this guy in the 1970s...", and, well...

Anyway, that makes a remarkable preamble to Ed Zitron: Make Fun of Them, which takes far too many words to get to the point which is that we need to start asking the "AI" proponents exactly what they're claiming. Ed points out that

Anthropic has now put out multiple stories suggesting that its generative AI will “blackmail” people as a means of stopping a user from turning off the system, something which is so obviously the company prompting its models to do so. Every member of the media covering this uncritically should feel ashamed of themselves.

Which, yes, is exactly the point of these stories: They're there to "humanize", to anthropomorphize, the LLM output. Because any remotely critical reading of this says that we should simply not give the random number generator access to the big red "blow shit up" button. But if we give these things some sort of agency in our minds, then we start to see what they're generating as somehow "intelligent".

This whole thing is feeling more and more like religion, with the evangelists talking about how amazing it is, and the rest of us sitting around saying "uh, what a bunch of self-referential easily disproven bullshit, and yet you keep sending me Bible quotes like they mean something..."

Anyway, yeah.