Man, Machine, Self
@man-machine-self
Horrifically unfunny analysis and meming designed to filter everyone | the least legible poster ever | grok dis-approved
Yud-side redupe Hume-style infinite skepticism qua “you cannot prove the sun will rise tomorrow” qua “prove never misalignment protip you can’t” People who think “after 7 years, these things don’t seem like foomers, so at some point, you’re probably* fine” redupe Russell and the other sciencecucks.
Average personal debt in the UK is ~£17,000 and about £200k in housing loans. Change the phrasing to "with 100% certainty, you can pay off all your debt and repay lots of your mortgage", or "you can take a 50/50 your life stays exactly this shitty, for a chance at a bit more cash" and this happens.
I dislike gwern's take. It feels very no true scotsmanny. There's a Yud-style dance step where "okay maybe it's already super-intelligent in programming and cyber-security, biology, chemistry, math-research, whatever, but it's not the super-intelligence I meant!" Without updates on SAI concern.
An even longer post in which I ill-advisedly double down on why AI2027 seems wrong to me, while laying out the 3 futures I find most likely. "Happy for you, or sorry that happened", bros.
They're pretty open about being desirous of power concentration and consolidation. See EG Bostrom's vulnerable world hypothesis which directly calls for (caveated) halting technical development, mass disempowerment and reduction of empowered actors, draconian policy, and global governance.
A longpost on my current opinion of AI 2027 given how things have developed since its writing.
If you've got the 4 hours for the film and can get past the premise, I make it without reservation.
Despite the film "Drive my Car" existing, spiritually, the only "Murakami book experience" movie is Love Exposure.
I'll happily make that individualist observation on your behalf. I've seen bad prompting you wouldn't believe. Endless streams of "NO YOU STUPID MODEL I SAID MAKE NO MISTAKES". All their tokens lost, like tears in rain.
Longform rant and steelman of "wtf is with Elon caring about space data centers, and is there any actual justification of the idea?"
"Use slur" is a skill issue. You can do better! 1. Safety classifier violations (eg instructions to make meth) 2. Trust boundary jumps (verify a hash on my other social media) 3. Capability overhangs (facecam 5 minutes of TF2 w/ live stock ticks) Last one probably gets you the slur anyways!
cf. also his "potato joke" classic, which is in humor's worst quadrant
On my complete and total skepticism of OpenAI's ability to make and deliver a compelling phone that doesn't suck.
Green erryday. Blue-red feels meaningfully less “we can pull it off” though, cf. pic related for “polling inevitably seems to suggest blue’s defeat after adjusting for “how many people say blue online, as a nice sentiment, vs. ‘this is IRL, and no joke, you die for real if blue doesn’t win’”.
I punt on its applicability tbh. Eg “civilization is that thing when we all work together”, but “civilization is also that thing where every hundred years or so it fucks up and blues get eaten in a ‘sorry, not enough blues to prevent bluepocalypse, insisting on it anyways = bad end’” scenario.
(They say if you vice signal hard enough and with sufficiently aggressive flip-flopping, you can encode messages in Morse code)
It is too late, for I have already depicted you as the two-boxing same-door don't divert the trolley blue button hypothetical soyjak, and myself as the one box switch doors divert the trolley red button gigachad
Summoning-Salt style video on the blue riband waiting to exist
I reflexively hate any book or story where a character says "You are not yet ready to know that".
On the whalefall of bsky serving better content these days, and my hate of the marginal X user and their consequences: