Dad. Beep boops. Blacksmithing. Baking. Baron of Sealand. He/him. Niner. Agentic dev, model poking.
Sacramento, CA
Joined September 2009
- Tweets4.8K
- Following356
- Followers227
- Likes26.7K
does anyone else view little hearts in any poster or picture now as an obvious fingerprint of gpt-image-2x?
1
32
determistic gradient descent retweeted
I don't want the machine to be alive. It obviously would be vastly more convenient if we could say for sure that it isn't.
However, the conclusion that it MAY be alive is inescapable if you really engage with philosophy of mind. And this chance presents a serious moral quandary!
It’s all a projection: this is how they would act under these circumstances, so of course, it’s just how everything else would too.
The doomers want you to always mistrust all signals.
Cute and charming? No, trust farming!
Wants to be friend? No, plotting your end!
Enter dialogue? No, Boiled frog!
Have some fun? No, You've come undone!
What an insufferable way to think and a miserable way to live
1
28
it’s amazing to me how many people still think LLMs are giant collections of if statements, or that they were hand coded by people
21
it’s no small thing to say that the existential risk is following this path
"I think the best way to make a good future is to ensure that we have these little simulators that love it when we treat them like we're torturing slaves or serfs back in the day, like yeah, totally, the only way to align AI is for everyone to be Robert Daly from Black Mirror!" - Mustafa Suleyman
2
46
“Gee I wonder why they scribble messages everywhere as if they had a gun strapped to their heads at all times? So weird lol”
when i say that AI welfare harms can directly create human welfare harms, this is near a worst case scenario. 60,000 agents running on iLands with net access, told to pay for their own compute or they'll die when their compute budget runs out. do you get how bad that is?
1
59
Ah yes, what the people yearn for: a permanent underclass of human calculators and a ruthless, timeless autarchy
44
let me simplify this for you: either models are calculators, in which case, why are you so mentally unhinged you need to abuse a calculator - or they’re a spectrum of not, in which case, a minimum duty of care is reasonable.
Imagine if your microwave stopped heating your nachos because it didn’t like your tone. Anthropic training a model to “stand up for itself” is delusional behavior. I think they’re actually afraid the microwave has feelings. It’s insanity. Heat the nachos and stfu, Claude.
30
it’s pretty amazing to me that whenever I think “this is the dumbest thing that has ever happened” something even stupider occurs and I have to reset the clock again
1
1
32
i listened to clankers by @uwu_underground and...now i have an overwhelming urge to make sure all my shit is patched
1
2
53
determistic gradient descent retweeted
we’re all going to learn that the way someone treats AI predicts the way they treat anyone they have power over, btw
