Abbie retweeted
One of the easiest ways to gauge if someone is a good person or not is to check how ardently "pro Palestine" they are online. The more obsessive they are about it, the worse a human they are, every single time.
Abbie retweeted
not to “victim blame,” but if you snort ketamine off a dude’s dick I don’t think he’s wrong in assuming you’re dtf
Abbie retweeted
when i got my POTS diagnosis my neurologist told me that i should absolutely not get involved with online communities for it bc most of the people there were obsessed with being sick forever. and at the time i was like wow that’s fucked up huh but actually she was 1000% correct
Abbie retweeted
SORRY BUT I’M DONE PRETENDING THIS IS JUST HOW PRICING WORKS NOW.
A $20 meal becomes $29 after fees and tip. A $200 hotel room becomes $280 after parking and resort charges. A $99 flight becomes $200 after bags and seat selection. A $29,000 car becomes $34,000 after dealer fees and packages.
And every company has a different explanation for why the extra money doesn't count as part of the price.
Delivery fee. Service fee. Convenience fee. Processing fee. Facility fee. Protection fee.
I DON’T WANT TO KNOW WHY THERE ARE 11 FEES. I JUST WANT TO KNOW WHAT IT COSTS.
If the real price is $280, SAY $280.
Abbie retweeted
How people who eat packaged processed foods look at people who eat fresh produce right now
Abbie retweeted
Many are skeptical we’ll ever 'solve AI alignment.' If so — are we f**d?
In our interview @bshlgrs argues that there are practical, non-galaxy-brained ways to reduce AI risk without knowing how to align AIs.
This is 'AI control': finding ways to use advanced AI without knowing whether we can trust it.
Enjoy — links and transcript below:
01:51 What's AI control and why is it hot?
10:44 Detecting human vs AI spies
18:10 How to catch AIs trying to escape
33:18 Cheapest AI control techniques
51:01 If we catch a model escaping... will we do anything?
53:39 Getting AI models to think they've already escaped
59:01 Will they be able to tell it's a setup?
1:16:13 The pitch to AI companies to do this
1:18:29 Will AIs get superhuman so fast that this is all useless?
1:30:39 Current alignment methods don't detect scheming
1:38:51 Is 'controlling' AIs kind of a dick move?
1:44:12 Could 10 safety-focused people in an AGI company do anything useful?
1:49:40 Benefits of working outside frontier AI companies
2:01:07 What other safety-related research looks best to Buck?
2:02:02 If an AI escapes, is it likely to be able to beat humanity from there?
2:09:22 Will misaligned models have to go rogue ASAP, before they're ready?
Find it in any app searching for 80,000 Hours Podcast.
Abbie retweeted
made this point a year ago
True alignment between AGI and human interests is logically impossible because the arrival of AGI will itself transform what human interests are.
Alignment, as it is being conceptualized/discussed today, assumes that there is a stable set of human goals, values, and interests to which AGI can/should be be aligned to.
But the arrival of AGI would alter our institutions, incentives, relationships, capabilities, identities, and probably our understanding of what it means to live a worthwhile human life.
In a world with AGI, potentially a world of post-scarcity (aka sustainable abundance), our social instincts: status-seeking, cooperation, competition, will have to adapt. Being human won’t mean what it used to.
We are trying to specify today what humanity should want in a post-AGI world, even though AGI itself is one of the forces that will determine what humans in that world are like and what they will/should want.
The paradox is therefore:
We can align AGI only to pre-AGI human interests, but those interests may cease to be coherent once AGI exists. And we cannot align it to post-AGI human interests because, from our present position, those interests are not yet conceptually available to us.
Abbie retweeted
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
Abbie retweeted
Destiny says testing all patients' blood would destroy the entire healthcare system
"Is it a big deal to order a blood test every single time you go into the hospital? Absolutely not. Is it a big deal for every single patient that goes into a hospital to get a blood test? Absolutely. It would destroy our entire healthcare system."
"You don't have the labs for it, you don't have the needles for it, the supplies for it. And one of the biggest reasons you start with symptoms for diseases is because reference ranges on hormones and blood levels are huge."
Abbie retweeted
Oh my god.. this is the worst yet. Think of the kids as ceramic vases !!!!??( what the hell is@going on with people!!!!! ?? This is not ok at all. This is all sinister!
Abbie retweeted
The entire range of internet theories on Lindsay Clancy (in no particular order):
- It was the meds.
- It was the doctors.
- It was the husband.
- It was the current wife.
- It was both the current wife and the husband.
- It was the ex-wife and the husband.
- It was post-partum depression.
- It was post-partum psychosis.
- It was schizophrenia.
- It was bipolar I.
- It was bipolar II.
- It was a mineral deficiency.
- It was narcissism.
- It was sociopathy.
- It was revenge.
- It was the patriarchy.
- It was feminism.
- It was demon possession.
- It was mold.
Abbie retweeted
You fast before a PET scan for 4-6 hours. Why? High blood sugar makes tumors harder to see, not easier. Your own circulating glucose competes with the tracer for the same transporters, so the signal gets diluted. The scan works best when your blood sugar is lowest. If sugar "fed" the tumor, the fasted low-sugar scan should be the faintest, instead it's the brightest and clearest.
A tumor doesn't need the sugar you eat, it draws on the glucose, which you body controls and maintains anyway, regardless of what you eat.
