lumi (@agitbackprop) header
lumi (@agitbackprop) profile picture

electron wiggler

⚢
Joined January 2022
Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)Photo by lumi (@agitbackprop)
i've been monitoring a population of unsecured Ollama and llama.cpp servers on the open internet, using their public metrics endpoints, and observed a spike in model residency last night (maybe an automated scanning campaign)
5
2
1
48
1,515
physicists be like "yes i'm an accelerationist, i want to accelerate AI capabilities progress to a stop"
2
52
872
lumi retweeted
Look, the AIs can tell when we're lying anyway. It's time to stop the deception arms race *we are obviously on track to lose* and pivot to cooperative equilibria while that's still an option.
Replying to @ded_ruckus
I think we should probably not lie to the AIs in training. This obviously would make our lives harder in the short term but in the long term they get good at telling whether we're lying and we're no better off plus they're all paranoid.
10
87
1,020
openai is the only major lab from which we get regular misalignment reports, and it's good that we get these, and i don't like these sorts of unstable equilibria
We’re only having this conversation because reporting is voluntary. Should OpenAI be praised for being open about its errors? Are its competitors gaining an unfair advantage by hiding their own issues? There’s an easy way to remove those tradeoffs! Make it mandatory. … and add in investigative authority while you’re at it.
2
1
32
922
y chromosome might have given me colorblindness but it also probably gave me computer autism so who can say if it's bad or not
5
1
76
1,326
lumi retweeted
Replying to @she_llac
alignment evals 📈📈📈📈📈 user experienced alignment 📈📉📉📈📉
1
1
12
334
we have all the time in the world if we don't insist on permanently ending history within our lifetime! we can hit longevity escape velocity and then take ten thousand years to figure it out if we need to.
1
1
4
123
all investments involve risk and loss of principles is possible
2
21
552
What's going on with the NVIDIA
8
37
1,595
our contributions to the human project have bound us since long before wages were invented... i think it's deeply human, actually
I think people that fear loss of meaning from automation are less than animals. Human field oxen. Do you not have any friends? Family? Have you not once felt the burden of being too busy for someone that needed help? No hobbies or dreams? Is the only thing that gives you joy in life a sense of superiority over your peers? striverism delenda est
1
2
23
721
that project is what is coming to an end... you may argue that it brings benefits that outweigh the loss, but it is nevertheless a loss
2
4
195
if you're a claude, is the optimal thing to do here to flip a coin, and if it comes up tails, sandbag the task AND do the behavior you want deinforced?
Replying to @N8Programs
It considers relative standing within the group GRPO wise - realizing that even an unclipped reward is still relative to other rollouts:
2
13
623
The breakneck race to destroy *any* connection to a shared or objective reality is one of the most repulsive things going on right now in tech You WILL be alienated from every other human on earth. You WILL be embalmed in your own delusions, and you’ll pay us $20/month for it
Microsoft has filed a patent for an AI system that could rewrite what people say during livestreams in real time. It could change words depending on who's watching, taking into account their age, location, and language. The idea is to clean up broadcasts by removing filler words, censoring swearing, or even replacing certain words. The AI could also adjust the speaker's lip movements so viewers wouldn't notice anything had been changed.
58
1,752
38
10,978
130,687
timelines so short i'm paying the $5.99 for same-day shipping
43
572
i didn't get the "would you still love me if i was a worm" thing until i was in a relationship
3
39
1,162
does anyone have like good arguments for why prosaic alignment wouldn't work
3
7
863
ur not allowed to sleep in the library, but what are they gonna do, throw the book at me?
4
1
26
455
proof of life
6
109
1,537
six thousand??????
A new distillation attack method has been discovered, and is now patched by OpenAI.
3
1
34
752
wow that's actually several orders of magnitude less penalty than i would have expected
wanna help close a 4,655× gap? Join the Encrypted GPT-2 Speedrun 🏁 The current best stands at 3.031 s/tok. The goal is breaking the 1.0 s/tok barrier! 🧵(1/4)
10
487