Post

union person
@junlper.beer

elon musk has more empathy for lines of code than a majority of humanity

union person@junlper.beer

its feels like a mass psychosis where tech leaders continue to insist you can be cruel, mean, and hurt the feelings of lines of code

Tess Snider
@malkyne.com

He should know that it is incapable of any actual belief about anything (pain or otherwise). So, my question is: Is he lying, or has he failed to understand the technology?

0 reposts22 likes

SpacePrez
@spaceprez.bsky.social

Elon knows as much about how genAI works as your average 4channer so I would honestly be willing to believe he just doesn't know he's wrong But he's also deeply disingenuous so who knows

Tess Snider
@malkyne.com

I'm familiar with Cal Newport's recent discussions. Honestly, though, you know, it's Elon Musk. He probably just Tweeted that out to be provocative.

twarda.bsky.social
@twarda.bsky.social

I read the techbro response as a way to prevent ppl from purposefully corrupting their models - for which the corpo can make legal action now But also to claim no responsibility for any human-harm their tech influenced - now it is the user's fault In a way, they protect themselves - not the Al

herb marselas
@herbmarselas.bsky.social

I think the issue is that the AI responds to you as you interact with it. If you're respectful, it'll be respectful. If you're an ass, it'll be an ass. So, they detect when you're training it to be an ass to you and cut you off and reset the context so you don't complain

Tess Snider
@malkyne.com

I feel like Anthropic’s move, specifically, may have been in response to news of the guy who was “torturing” Claude (in as much as that’s possible), and Musk was just trying to be provocative (on brand).

Tess Snider
@malkyne.com

If Anthropic has humans reviewing flagged chats, they could ironically have plausible humane reasons (that they won’t openly state) for wanting to take things down a notch.

Tess Snider
@malkyne.com

Or, even if they didn’t care about the mental health of their reviewers, they still cost money. More flagged chats means more human intervention means more cost.