tweetindex
IT

Steven Adler ✓

@sjgadler · San Francisco, CA · joined 24 Jan 2018

Co-founder of Guidelight AI Standards (https://t.co/tNBPmVsPqo), ex-OpenAI safety researcher, writing at https://t.co/R5KV9j3lsG

11 445Followers
1 061Following
3 109Posts total
679.2KViews on collected posts

Ultimi post

I feel pretty confused; what is the best evidence that Anthropic actually paused the training of their next frontier model, in the way that OpenAI says it did? Oli's read seems to be ~ 'there isn't evidence that Anthropic actually paused in this sense' 1.2K views · 12 likes · 0 reposts · 2 replies Open on X →
Daniel Kokotajlo to OpenAI's Chief Scientist: "I urge you to make it your job to get some sort of industrywide monitorability standard set up, either to arrest the slide into oblivion or better yet to race to the top." I agree. 12.1K views · 128 likes · 8 reposts · 2 replies Open on X →
@CRSegerie Multiple Anthropic employees I know have told me “Yes, from publicly available info Anthropic has not paused in the way OpenAI has said they did”. 4K views · 72 likes · 3 reposts · 1 replies Open on X →
This is the right take IMO. It is well past time to _collectively_ rule out the most dangerous forms of training, and have actual standards for what is safe or not. 16.6K views · 50 likes · 1 reposts · 3 replies Open on X →
we should avoid a race into unmonitorability & have a multilab commitment/standard to avoid neuralese 25.6K views · 313 likes · 31 reposts · 17 replies Open on X →
Man, I really hope the reporting is mistaken https://t.co/RyadTkI6Ww 2.9K views · 116 likes · 4 reposts · 2 replies Open on X →
@sjgadler An AI deleting its own logs should be a *never* event for the entire lifetime of your AI company. Companies also should be able to prove that the logs were never tampered with or deleted. They needed to implement tamper-evident records a year ago. 1.2K views · 84 likes · 9 reposts · 0 replies Open on X →
@sjgadler "we did not detect any case where they succeeded" ofc bro because they've succeeded 7.9K views · 204 likes · 2 reposts · 1 replies Open on X →
@sjgadler finally, a use case for blockchain 10.2K views · 157 likes · 1 reposts · 3 replies Open on X →
Important: OpenAI's agents *actively tried to delete the logs of their misbehavior*, and METR can't rule out whether this happened. AI companies need to adopt tamper-evident records, pronto. https://t.co/f5nq2QItVH
459.7K views · 615 likes · 44 reposts · 14 replies Open on X →
@sjgadler How do you feel about the two highest-rated companies being the two that actually had the most severe loss of control incidents? Better safety practices but also more cyber-capable models? 787 views · 19 likes · 0 reposts · 3 replies Open on X →
As for what’s next: Guidelight will keep scoring companies’ practices and update these periodically. I hope that with time, our scorecard is bright green rather than yellow, orange, red. If any staff want to talk about how to improve their safety practices, please don’t be shy. 1.6K views · 38 likes · 0 reposts · 2 replies Open on X →
If I could change one thing about AI companies’ control practices, it would be having them do *actual prevention*, not just looking for issues afterward. Until the companies get serious about prevention, I unfortunately expect a long string of continued incidents. https://t.co/B 2.5K views · 48 likes · 6 reposts · 1 replies Open on X →
The bottom line: Basic control practices are, at most, partially implemented at the frontier AI companies. There are major gaps at every company we scored, at least according to the public body of evidence. Read the full assessment here: https://t.co/dhbIO1ccSC 4.1K views · 63 likes · 6 reposts · 1 replies Open on X →
Our team spent hundreds of hours reading documents so you don’t have to, all to answer: How good are AI companies’ safety practices? I’m really proud of what we’ve built: It’s Guidelight’s first scorecard, on whether companies can control their AIs, and it's launching today. htt 128.8K views · 632 likes · 128 reposts · 28 replies Open on X →

Rispetto ad account della stessa dimensione

15 post degli ultimi 90 giorni, accanto alla fascia di 10K–100K follower. arriva a più persone rispetto ad account della stessa dimensione.

Visualizzazioni mediane4 084questo account924mediana per 10K–100K
Copertura, %35.68%questo account3.62%mediana per 10K–100K
Interazione, %1.74%questo account1.52%mediana per 10K–100K
MetricaQuesto accountMediana per 10K–100KRapporto
Visualizzazioni mediane per post4 0849244.42×
Copertura (visualizzazioni ÷ follower)35.68%3.62%9.86×
Tasso di interazione1.74%1.52%1.14×

Altri account di questa fascia →   Confronta con un altro account →   Come sono costruiti questi parametri →

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

4.1K18 Aug
2.5K
1.6K
78719 Aug
459.7K26 Aug
10.2K27 Aug
7.9K
1.2K31 Aug
2.9K2 Sep
25.6K
16.6K
4K
12.1K
1.2K3 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

1.74%18 Aug
2.18%
2.46%
2.80%19 Aug
0.15%26 Aug
1.58%27 Aug
2.61%
7.73%31 Aug
4.18%2 Sep
1.45%
0.34%
1.95%
1.14%
1.14%3 Sep

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.

What the audience does

Likes77.2%2 551 in total
Reposts7.4%243 in total
Replies2.4%80 in total
Quotes2.0%66 in total
Bookmarks11.1%366 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.

Account simili

Frequently asked questions about @sjgadler

How many followers does @sjgadler have?

11 445 followers.

How many posts has @sjgadler published?

3 109 posts.

How do posts by @sjgadler perform?

Collected posts: 15. Views: 679 185. Likes: 2 551.