This is a special post for quick takes by a3w. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Listened to "its not just a fucking sandbox" on AI article readings. My takeaways:
my following bias got affirmed: Xitter is only Nazis and trolls
They argued: "I am a human being and want to have spare time, not get attacked myself" is fair but not interesting. See 1 as to my expectations for social media.
And as for why models escape: It's not just the fucking sandbox, the whole company (is the problem as it) fights "complexity" and "scaling", so they are doing a good job, but should not do the job at all is the right solution to save humanity. (Another path would be to hold AI firms accountable for actions of their models. If you look for negligence and only find the reason of "making money is more important than scaling down complexity and missing out on revenue due to not scaling up so fast", then capitalism and company are both at fault, but you cannot sue capitalism, so 100 percent of accountability is back at OpenAI, IMHO.