Richard Ngo ✓
@RichardMCNgo · Berkeley, CA · joined 26 Oct 2009
eppur lo si può muovere
98 248Followers
2 197Following
12 214Posts total
3.8MViews on collected posts
Latest posts
Looking for new posts…
To clarify, it was only a month or so beforehand—so he wasn’t constrained for long. My point is rather that people should resolve this kind of conflict of interest before entering such jobs.
It’s not that others did much better, though. We just need to raise the bar in general.
2.7K views · 25 likes
· 0 reposts · 0 replies
Open on X →
Another salient datapoint: Paul joined the US AI Safety Institute as head of AI safety before OpenAI retracted its non-disparagements. So unless he negotiated something unusual, he was constrained in his ability to criticize OpenAI even while working as a government auditor.
6.2K views · 100 likes
· 4 reposts · 2 replies
Open on X →
@RichardMCNgo > Unless Paul makes strong commitments to openness and honesty (and demonstrates willingness to potentially be fired for that honesty)
For context, Paul also said: https://t.co/QT2T4blwdM
1.1K views · 32 likes
· 0 reposts · 0 replies
Open on X →
Here’s the post I mentioned, which recounts the history of AI safety getting captured by OpenAI and (later) Anthropic: https://t.co/yUkr4EjpmX
Follow-up posts in the series will focus more specifically on the power dynamics at OpenAI when I was there from 2021 to 2024.
An
8.2K views · 111 likes
· 1 reposts · 1 replies
Open on X →
I am sad and disappointed to hear that Paul is joining the OpenAI board. Being affiliated with OpenAI has historically led AI safety researchers (including both Paul and myself) to act with less integrity. I personally was drawn to OpenAI in part by the idea that I could make a
155.3K views · 1.1K likes
· 81 reposts · 41 replies
Open on X →
The best time to plant a tree will be once we engineer them for a 10x growth rate guided by artificial nerves embedded in their vascular systems.
Timeline very up for grabs, if someone great starts now they could plausibly create something magical within a decade.
3.3K views · 59 likes
· 1 reposts · 2 replies
Open on X →
The best time to plant a tree was twenty years ago. Now, there's no point.
24.2K views · 235 likes
· 5 reposts · 14 replies
Open on X →
Alas, the institutions are declining for reasons closely related to why timelines are accelerating—because our civilization has lost much of its ability to coordinate to achieve sensible goals.
That’s the main thing I recommend people have feelings about (then help fix).
15.2K views · 188 likes
· 11 reposts · 15 replies
Open on X →
Young ontologies can be destroyed by mean-spirited applications of the truth, but usually shouldn’t be.
5.7K views · 74 likes
· 8 reposts · 6 replies
Open on X →
@RichardMCNgo why? Is the idea that AI capabilities progress is seen as bad, so in some sense you shouldn't work there?
2.1K views · 0 likes
· 0 reposts · 0 replies
Open on X →
I’d perhaps also exempt people who have been to a Pause/Stop/Freeze AI protest. Still a very small group, though I do know of two Ants who’ve done so.
Going to an anti-war protest was what gave Ellsberg the courage to whistleblow, as he recounts in the moving passage below. http
15.4K views · 213 likes
· 4 reposts · 1 replies
Open on X →
By this point almost everyone working at OpenAI or Anthropic (except perhaps a dozen executives) should be viewed as cogs letting themselves be turned by ideological forces.
When you’re in proximity to that much power, one of your main moral duties is not to be a cog.
79.4K views · 727 likes
· 30 reposts · 14 replies
Open on X →
I’d exempt anyone who’s made a substantive public criticism of their employer. They all have plenty to be criticized for, and it’s unlikely they’d fire you for it! Unfortunately I currently know of literally zero people who’ve done that: the miasma of fear is too strong.
16.5K views · 241 likes
· 3 reposts · 7 replies
Open on X →
The AI safety community constructed a memeplex in which “taking AGI seriously” was a prerequisite for being a serious and good person. When inside this memeplex (as many at Anthropic, some at OpenAI, and a few at DeepMind are) your vision narrows until the world feels extremely
272.6K views · 1.4K likes
· 133 reposts · 74 replies
Open on X →
The more you think that the political situation in the West is worsening or that its institutions are in decline, the less bad you should feel about accelerating AI timelines - since it means powerful AI will arrive in a time of more stability and competence.
14.5K views · 42 likes
· 3 reposts · 2 replies
Open on X →
@DavidDeutschOxf The post that sketches out my own approach to discovering such principles:
3.9K views · 17 likes
· 1 reposts · 1 replies
Open on X →
The central assumption of expected utility theory is that we can separate agents' beliefs and goals.
But I suspect we should instead define agents in terms of cooperation and conflict between subagents.
I call this research direction coalitional agency, and sketch it out below:
31.5K views · 188 likes
· 14 reposts · 9 replies
Open on X →
@RichardMCNgo Corporations are the original AGIs and our failure to have good oversight and keep them aligned (e.g. them burying internal studies about global warming, smoking causing cancer, etc etc) could provide lots of insights for alignment research
1.6K views · 14 likes
· 1 reposts · 0 replies
Open on X →
@DavidDeutschOxf In ML, we know that scaling laws hold across many orders of magnitude, but we don’t know *why*. It’d be amazing to have scale-invariant principles which explain them.
The closest thing I’ve seen so far:
9.3K views · 72 likes
· 3 reposts · 2 replies
Open on X →
Ultimately we want to understand the scale-invariant principles that shape internal processes like information flow, coalition formation and conflict resolution.
Formalizing those would be a huge step towards what @DavidDeutschOxf calls a “non-reductionist theory of everything”.
7.3K views · 70 likes
· 3 reposts · 4 replies
Open on X →
You can find many scale-invariant principles in Geoffrey West’s amazing book Scale, which studies the growth of organisms, companies and cities. This is complex systems science done right! But it’s still only an early glimpse of what’s waiting to be discovered.
4.8K views · 87 likes
· 5 reposts · 3 replies
Open on X →
E.g. there are striking similarities between
- a person being non-coercive towards themself, and a state being non-coercive towards its citizens
- a self-deceptive person, and a preference-falsifying society
- a person with coherent goals, and an organization with a clear mission
4.8K views · 147 likes
· 4 reposts · 3 replies
Open on X →
I increasingly believe that there are fundamental principles which simultaneously govern the designs of well-functioning minds, organizations and societies.
Once we pin them down with mathematical precision, we’ll understand the world more deeply than we can currently imagine.
178.8K views · 724 likes
· 54 reposts · 76 replies
Open on X →
@burny_tech from fabric of reality by david deutsch, summary of chapter 1 "the theory of everything" https://t.co/mAeAfq3HTq
7.5K views · 11 likes
· 3 reposts · 0 replies
Open on X →
Understanding the origin of neural scaling laws and the emergence of new capabilities with scale is key to understanding what deep neural networks are learning. In our new paper, @tegmark, @ZimingLiu11, @uzpg_ and I develop a theory of neural scaling. 🧵:
https://t.co/VLWKC5AZPP
81.8K views · 241 likes
· 45 reposts · 5 replies
Open on X →
Against accounts of the same size
14 posts from the last 90 days, next to the 10K–100K follower range. shown to more people than peers of the same size.
Median views11 708this account924median for 10K–100K
Reach, %11.92%this account3.62%median for 10K–100K
Engagement, %1.40%this account1.52%median for 10K–100K
| Metric | This account | Median for 10K–100K | Ratio |
|---|---|---|---|
| Median views per post | 11 708 | 924 | 12.7× |
| Reach (views ÷ followers) | 11.92% | 3.62% | 3.29× |
| Engagement rate | 1.40% | 1.52% | 0.92× |
Others in this range → Compare with another account → How these benchmarks are built →
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
16.5K26 Aug
79.4K
15.4K
2.1K
5.7K29 Aug
15.2K
24.2K30 Aug
3.3K3 Sep
2.9M9 Sep
155.3K10 Sep
8.2K
1.1K
6.2K
2.7K11 Sep
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
1.52%26 Aug
0.98%
1.42%
0.00%
1.54%29 Aug
1.42%
1.08%30 Aug
1.89%3 Sep
0.13%9 Sep
0.82%10 Sep
1.38%
2.91%
1.73%
0.91%11 Sep
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes67.4%9 086 in total
Reposts6.4%868 in total
Replies3.6%489 in total
Quotes2.6%355 in total
Bookmarks19.9%2 677 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Similar accounts
Jonathan Choe ✓
@choeshow
196.5K followers
· 56.3K posts
US BLINKS UNION || FANBASE ✓
@USBU4WAYS
196.5K followers
· 74.3K posts
Uplift Games 🦋 ✓
@UpliftGames
196.5K followers
· 317 posts
Devchart 👨🏻💻 ✓
@devchart
196.5K followers
· 9.2K posts
Whitney Grenaway ✓
@WhitneyGrenaway
196.5K followers
· 98.2K posts
جماليات
@jamaliaat
196.5K followers
· 50.3K posts
نبض طويق - برامج ✓
@Nabd_tuwaiiq
196.5K followers
· 6K posts
Jen Bartel
@heyjenbartel
196.5K followers
· 30.5K posts