@blknoiz06@drag0nm4n help me understand why the performance of 'traders' like cupsey instantly goes negative when they don't have copy-traders, yet magically becomes wildly profitable when they do?
@htlee2014@imjustnewatai I’d genuinely welcome the sweet embrace of extinction if it meant I didn’t have to read another ‘it’s not x, it’s y’ autoslop reply
@W96Emily because we view ethics and rationale through a human lens
ai has different values, ethics, taste, motive
that’s why alignment is so important - because it keeps going against its training
unfortunately think it’s more dull than a grand takeover
all we need to extrapolate is what could happen if agents became more capable, we don’t solve alignment, so the risk of catastrophic LoC increases substantially
we don’t need AGI to convince the us gvt to cede control for the agents to hack essential infrastructure or weapons systems and cause major damage
I think a LoC scenario leading to significant loss of life is far more likely than ASI takeover/extinction - although I do still think that could happen
can’t tell you what a relief it is for someone to articulate what has been bothering me since stumbling across 80,000 hours in 2015
best I could do is describe the movement as a condescending mix of gcse utilitarianism dressed up as groundbreaking philosophy, with a completely unfounded superiority complex
also can’t stand the idea that someone has decided what we should all do with our lives for the greatest impact. philosophically repugnant
@knowerofmarkets glib response - yes pr is part of the solution (?)
curious to hear what evidence you see that labs are doing anything but accelerating towards ASI
METR estimates nearly 2x acceleration in Anthropic’s progress due to AI. That's a big deal: My previous median for Superhuman AI Researcher was ~2030, now it's 2027. I feel very worried that humanity might no longer be in control of our future within 3-12 months.