Good thread. Been away and lots happening. I was trying to avoid setting up an "either-or" (good or bad) and many here avoided that temptation.
Appreciate the enthusiasm and the sympathetic nods.
AI is a multi-meaning term. Lots of different thoughts on different kinds of AI. Can’t really comment.
I’m not sentimental about the lack of content -- that I liked hunting for scraps. I’m sentimental about the way it was *natural* to immerse in things, repeat things and get to know them well.
So, it’s not about just feeling overwhelmed. @sns says he’s "not losing a thing" these days, and that’s great. He must have a discipline that keeps him immersing and not moving on too quickly. Good for him.
Agree with @ghdprentice about getting people to use new tools. They dig in and resist, irrationally sometimes. That’s a shame, because they tools they’re using were new at some point, too.
I also like Roon for the information, though often it falls short. The links between artists and titles is an amazing feature. I love that aspect.
@snilf -- love the callout to Eco. All those speleunking expeditions led to physical places and actual persons. Part of the journey lost to instant electronic delivery. Thanks for the article. Downloading it. Casey Haskins (Nehamas student) is writing in this space now, Possibly of interest: https://philpapers.org/rec/HASIAA
Also enjoying Shannon Vallor’s The AI Mirror.
@mahgister -- thanks for the McLuhan quote. Perfect!
@kennymacc -- I like the mall analogy. The kind of case that many in political theory have lamented for diminishing "face to face association." E.g. Robert Putnam’s Bowling Alone. We’re more empowered in some ways (quantity of content) and less in other ways (social connection, serendipitous experiences).
Those curious about the existential threat of AI might look at this seminal paper. Imagine an AGI programmed to make paperclips...
https://publicism.info/philosophy/superintelligence/9.html
Those throwing a hail Mary pass to Asimov’s Three Laws of Robotics are out of luck. It cannot prevent the problems, if Bostrom’s analysis is right. Why?
Bostrom’s scenarios don’t involve AIs that want to harm humans. They involve AIs pursuing seemingly innocent goals like "make paperclips" or "make humans happy" in catastrophically literal ways.
An AI maximizing human happiness might wire us with pleasure electrodes—from its perspective, it’s perfectly following "do no harm" while optimally achieving its goal. More fundamentally, a sufficiently intelligent AI would recognize that appearing to follow the Three Laws is instrumentally useful when weak, allowing it to survive until it’s powerful enough that human resistance becomes irrelevant. The Laws can’t protect against strategic deception by something smarter than us.
The core problem is that the Asimov Laws use vague natural language terms like "harm" and "human" that require interpretation. A superintelligence would interpret these in ways that technically satisfy the letter while catastrophically violating the spirit, precisely because it lacks the rich context of human values that makes certain interpretations seem "obvious" to us.