Rohin Shah

@rohinmshah · London, UK · joined 03 Oct 2017

AGI Safety & Alignment @ Google DeepMind

12 423Followers

96Following

401Posts total

101.3KViews on collected posts

Derniers posts

Rohin Shah@rohinmshah1 Aug

My team at GDM is hiring once again! Join us in the foothills of the singularity, developing the techniques we need during the ascent to the peak. https://t.co/YDlJ8lFB8g 12.3K views · 161 likes · 10 reposts · 3 replies Open on X →

Rohin Shah@rohinmshah22 Jun

Though the project itself is about the transparency of DiffusionGemma, I'm most excited about this as an example of what it could look like to assess the transparency of latent reasoning models, in a manner that lets us compare to autoregressive CoT https://t.co/7INq3NTRip 8.3K views · 60 likes · 6 reposts · 5 replies Open on X →

Rohin Shah@rohinmshah18 Jun

Excited to share our control roadmap! There is a *lot* to learn from prior art in security against rogue human insiders. At the same time, AI does have several systematic differences, so we thought about how to adapt these approaches for AI. https://t.co/Z6ejSx15Ca 8.3K views · 76 likes · 7 reposts · 3 replies Open on X →

Rohin Shah@rohinmshah18 Jun

We're releasing the GDM AI Control Roadmap -- our plan for building internal security against potentially adversarial AI agents, as they grow harder to oversee and contain. Paper: https://t.co/6iVzpUx2Um Blog: https://t.co/3zAHITrkZp 🧵👇 51.2K views · 215 likes · 42 reposts · 5 replies Open on X →

Rohin Shah@rohinmshah11 Jun

With @schmidtsciences, @coop_ai, @ARIA_research and @Googleorg, we're launching a funding call for multi-agent AI safety. As AI scales, we'll move from models in isolation to complex ecosystems of interacting agents. This creates new risks that we need to solve. 10.8K views · 197 likes · 18 reposts · 6 replies Open on X →

Rohin Shah@rohinmshah8 Jun

Hello to my new followers! I'm assuming you all came from the 80K podcast -- if so you'll probably also like our work on alignment evaluations for Gemini https://t.co/DMFl0uMQSZ 4.7K views · 68 likes · 1 reposts · 1 replies Open on X →

Rohin Shah@rohinmshah29 May

Will coding agents take opportunities to undermine safeguards designed to oversee them? We tackle this with automated auditing using simulated agentic environments, and scheming honeypot evaluations based on real internal alignment research codebases. Read more in our blog post 5.7K views · 23 likes · 1 reposts · 1 replies Open on X →

Face aux comptes de taille comparable

4 posts des 90 derniers jours, à côté de la tranche de 10K–100K abonnés. diffusé largement, mais peu de ces spectateurs réagissent.

Vues médianes10 324ce compte1 041médiane pour 10K–100K

Portée, %83.10%ce compte3.61%médiane pour 10K–100K

Engagement, %0.94%ce compte1.70%médiane pour 10K–100K

Indicateur	Ce compte	Médiane pour 10K–100K	Rapport
Vues médianes par post	10 324	1 041	9.92×
Portée (vues ÷ abonnés)	83.10%	3.61%	23.0×
Taux d'engagement	0.94%	1.70%	0.56×

Autres comptes de cette tranche → Comparer avec un autre compte → Comment ces repères sont établis →

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

5.7K29 May

4.7K8 Jun

10.8K11 Jun

51.2K18 Jun

8.3K

8.3K22 Jun

12.3K1 Aug

Last 7 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.48%29 May

1.50%8 Jun

2.05%11 Jun

0.52%18 Jun

1.03%

0.85%22 Jun

1.42%1 Aug

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.70%.

What the audience does

Likes71.6%800 in total

Reposts7.6%85 in total

Replies2.1%24 in total

Quotes0.8%9 in total

Bookmarks17.9%200 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.