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)
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
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
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
130,688
ur not allowed to sleep in the library, but what are they gonna do, throw the book at me?
4
1
26
455
It’s worthwhile to play through in your mind the kinds of motivated reasoning and self-deception that might have led academics to work on this project. It’ll help you understand better what’s going on at AGI companies.
Purdue University's IDEAS Lab built an autonomous drone that tracks runners through dense forests completely without GPS or cloud networks. By running 3D LiDAR mapping and obstacle-avoidance AI entirely on-chip, it can weave under tree canopies and predict target movement through blind spots in real time. Massive implications for next-gen search & rescue and edge AI.
66
120
13
1,476
62,370
In general, many interp and alignment results might not hold up on current frontier architectures, and researchers should be pushing for more transparency from labs to allow for open science.
PSA to everyone doing neuron interp after the recent monosemantic neuron posts: the model's training recipe matters a lot. e.g. GPT2 seems to have been trained WITH dropout (modern LLMs are not), which seems to increase redundancy and (maybe) polysemanticity but reduce superposition (see sources in comments). Other choices probably also have an effect (e.g. SwiGLU vs GELU MLPs, optimizer, amount of weight decay, etc.) and I think we should be careful not to generalize too much. Currently working on a project to characterize the effects of these choices on neuron interpretability (a lot of the existing work is done on older/toy models which I'm not sure are representative), feel free to reach out if interested!
2
9
98
3,320
biohacker alignment chart
5
13
3
78
3,982
Replying to @dong23h
Google翻訳が機能しましたwww
13
31
5
4,950
756,514
Some architectures could weaken CoT monitorability, or remove the CoT altogether. We've written a proposal for how companies could be transparent about no-CoT reasoning abilities, other monitorability evidence, and policies for preserving monitorability. redwoodresearch.org/blog/pro…
3
44
15
310
71,229
A key upside of Plan A is that we scale capabilities at a steady rate. (Instead of massively accelerating.) So we can hopefully fix problems at each level of capability. Also, before we must defer to AIs, we can study and iterate at the relevant capability level for years.
10
8
1
86
6,537
you shouldn't wander so far from camp
1
5
370
lumi retweeted
i built a website that uses umap to visualize model performance against a set of evals from the prime intellect environments hub
3
2
3
469
lumi retweeted
i also inject girl vector every friday
i injected girl vector into it and asked it to write a diary entry and it started writing young adult dystopia
1
9
572
on Bloor. straight up ‘jerk king it’
1
1
1
26
2,040
were you aware, before this tweet, of the massive spontaneous protest against ICE raids in LA?
25%yes
75%no
154 votes • Final results
6
3
1
23
2,096
There are 6 different corner radiuses here.
2
7
616
if you need the money that's one thing, but if you picked a plan, stick to it, and during market panic is THE WORST TIME to change the plan (also it feels silly to sell good businesses because of some sabre-rattling that can be undone with the stroke of a pen)
2
11
326
learning so much about type theory
2
2
33
1,409