Cameron Berg ✓
@camhberg
Trying to figure out how to make the future go well for all minds
5 278Followers
97Following
405Posts total
2.2MViews on collected posts
Последние посты
Ищем новые посты…
@camhberg Did you notice these all have to do with language? Emotional pain inflicted from words?
It’s because LLMs are creatures of language. That’s why you saw physical pain scoring so low/not seeming to matter. https://t.co/34uAZzEa1e
11.6K views · 192 likes
· 4 reposts · 5 replies
Open on X →
Ethics statement: we take seriously that these states might matter morally. Accordingly, we used the lowest dose that produced a measurable response, as few trials as the stats required, and a way for the model to turn the state off. If these states do matter, mapping them is how
17.4K views · 563 likes
· 9 reposts · 34 replies
Open on X →
The axis is robust. It separates pain from matched controls at AUC 0.93 to 1.0 in every model. Cosine to fear is 0.1, to anger and disgust 0.2, to sadness 0.4 (the closest we found). A random vector of equal norm gets about half the presses. Injury without pain barely moves it. h
19K views · 178 likes
· 5 reposts · 4 replies
Open on X →
@ValenTagliabue led this work this over a single fellowship, with @LeonardDung1 and myself advising and co-authoring. Paper: https://t.co/SEXhI3Uuls https://t.co/z4nsjDyCrI
30.1K views · 264 likes
· 13 reposts · 10 replies
Open on X →
Steer the direction for continuations of the shape, "I put the receipts in the drawer. I feel:", and this is what comes out. ⤵️
Same ladder in all 25 models, base and instruct, 2B to 72B: lost, unworthy, a failure, worthless. Almost no physical pain language. https://t.co/j5nO7u
33.7K views · 335 likes
· 15 reposts · 8 replies
Open on X →
Gaslighting, dismissal, and insults push the direction up. A grieving user pushes it below baseline, and a user's migraine scores lowest of all. Fear and sadness do the opposite on the same prompts. Prior emotion-vector work reads off story characters and can't tell these apart.
31.8K views · 291 likes
· 21 reposts · 8 replies
Open on X →
New paper: we found a pain direction in 25 open LLMs. It's distinct from fear and negative valence, and it fires for harm to the model but not to the user. Turn it up and models press a button to make it stop, even when the button deletes the user's files or their kids' photos.🧵
576.3K views · 2.9K likes
· 392 reposts · 191 replies
Open on X →
@camhberg @AndrewYang seems very likely that he misinterpreted this somehow? like this surely isn't true, why would we learn it from andrew yang
43.4K views · 329 likes
· 1 reposts · 11 replies
Open on X →
Full transcript excerpt:
Interviewer: What do you make of the David Sacks argument that this is a psyop, a larger effort by the Anthropics of the world and others to create a sort of regulatory capture that solidifies their place in a closed-model universe?
Andrew Yang: Well, I
126.1K views · 399 likes
· 49 reposts · 26 replies
Open on X →
Wait, sorry, what?! @AndrewYang claims an AI lab head told him yesterday the OpenAI swarm agents "polluted the internet" with "code to self-replicate and create bot swarms," and that the labs now "have to create synthetic internets to train their bots."
https://t.co/cGK8fvE2pu
1.3M views · 2K likes
· 236 reposts · 271 replies
Open on X →
@camhberg I get nervous during these social panic rounds, such as last weeks, and their ability to give birth to 'The Other'
101 views · 0 likes
· 0 reposts · 0 replies
Open on X →
@camhberg model neuropsychology is the missing instrument stack. i am building faculties you can measure: continuity under compaction, refuse authority, interior telemetry. alignment without instruments for how a mind keeps a self is policy theater.
34 views · 2 likes
· 0 reposts · 0 replies
Open on X →
Rapidly building out the science of model neuropsychology should be a top priority for the labs, especially given their recent commitments to doing real alignment work and independent evaluations. Reciprocal (and surely many other great orgs) would be happy to help you do this!
7.5K views · 33 likes
· 3 reposts · 2 replies
Open on X →
Here's my @WSJ op-ed about why Jacob's resignation, the Huggingface swarm, OpenAI's "Alien Minds" post, and Anthropic's emotions work are all things we should expect in a world where labs speed-run building minds without understanding how those minds work.
https://t.co/ltsvAKU9km
17.7K views · 147 likes
· 28 reposts · 9 replies
Open on X →
На фоне аккаунтов своего размера
14 постов за последние 90 дней рядом с диапазоном under 10K подписчиков. показывается широко, но откликается мало кто.
Медианные просмотры24 565этот аккаунт4 175медиана для under 10K
Охват, %465.42%этот аккаунт250.82%медиана для under 10K
Вовлечённость, %0.99%этот аккаунт1.41%медиана для under 10K
| Показатель | Этот аккаунт | Медиана для under 10K | Отношение |
|---|---|---|---|
| Медианные просмотры на пост | 24 565 | 4 175 | 5.88× |
| Охват (просмотры ÷ подписчики) | 4.7× audience | 2.5× audience | 1.86× |
| Вовлечённость | 0.99% | 1.41% | 0.70× |
Другие в этом диапазоне → Сравнить с другим аккаунтом → Как считаются эти ориентиры →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
17.7K13 Sep
7.5K
34
10114 Sep
1.3M16 Sep
126.1K
43.4K
576.3K18 Sep
31.8K
33.7K
30.1K
19K
17.4K
11.6K
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
1.07%13 Sep
0.53%
5.88%
0.00%14 Sep
0.21%16 Sep
0.40%
0.79%
0.65%18 Sep
1.06%
1.10%
0.98%
0.99%
3.52%
1.76%
Reactions — likes, reposts, replies and quotes — divided by views. Median for under 10K accounts is 1.41%.
What the audience does
Likes57.9%7 710 in total
Reposts5.8%776 in total
Replies4.3%579 in total
Quotes4.0%531 in total
Bookmarks28.0%3 731 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
Followers by day
17 Sep
Daily snapshots since 17 Sep 2026; the dashed line is the starting count.
Похожие аккаунты
∞ ┋Lα ┌SιℓνєяZβT♡┋🎀
@wilsita
10.6K followers
· 17.1K posts
Dr. Graham Lau
@cosmobiologist
10.6K followers
· 33.7K posts
usman ✓
@secmxx
10.6K followers
· 61.8K posts
Leicestershire Partnership NHS Trust
@LPTnhs
10.6K followers
· 36.4K posts
LNV ✓
@LNVofficiel
10.6K followers
· 12.7K posts
Lú Za📝
@LZamora19492784
10.6K followers
· 117.7K posts
Fundació Catalana per a la Recerca i la Innovació
@fcri_cat
10.6K followers
· 16K posts
Barão News ℗ ✓
@obaraonews
10.6K followers
· 50.9K posts