Puria retweetledi
"Per aspera ad astra" is so profoundly poetic. Through thorns, to the stars. All the way.
English
Puria
1.9K posts
Puria
@RadmardPuria
Cofounder & Director @GeodesResearch. Views are my own.
Cambridge, UK Katılım Kasım 2019
762 Takip Edilen288 Takipçiler
Puria retweetledi
Have a look at aisafety.com/map yall, I just realised not many people are aware of this masterpiece
Anastasiia Gaidashenko@avgaydashenko
Such a good list! I'd also add: - Astra Fellowship by @ConstellOrg - SPAR by @KairosAIS - LASR Labs - AI Safety Research Fellowship by @pivotal_org - Cambridge ERA:AI Fellowship (@era_cambridge) - Algoverse AI Safety Fellowship - PIBBSS - CHAI There's a host of non-technical fellowships as well, lmk if it'd be useful to compile such list
English
Puria retweetledi
It "feels like the first smart model in a long while" due to this
MetaCritic Capital@MetacriticCap
Opus 4.8 is the first smart model in a long while
English
Puria retweetledi
We've raised $65 billion in Series H funding at a $965 billion post-money valuation, led by @AltimeterCap, Dragoneer, @Greenoaks, and @sequoia.
This investment will help us advance our research and expand our capacity to meet growing demand for Claude.
English
Puria retweetledi
Thanks to a generous philanthropic grant from @coeff_giving (pending final logistics), 𝘎𝘦𝘰𝘥𝘦𝘴𝘪𝘤 𝘪𝘴 𝘩𝘪𝘳𝘪𝘯𝘨 𝘧𝘰𝘶𝘳 𝘔𝘦𝘮𝘣𝘦𝘳𝘴 𝘰𝘧 𝘛𝘦𝘤𝘩𝘯𝘪𝘤𝘢𝘭 𝘚𝘵𝘢𝘧𝘧. Come build the base of alignment with us 🤖
We're a Cambridge-based AIS org. Our seminal work (alignmentpretraining.ai) showed you can bake alignment priors into base models.
Applications now open: airtable.com/appuugUGFPJEy6…
English
Puria retweetledi
Puria retweetledi
It's gonna be the last summer of the Anthropocene. Enjoy it.
English
Puria retweetledi
For my non-tech friends, this is like Ronaldo joining Manchester City
Andrej Karpathy@karpathy
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
English
Puria retweetledi
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
English
Puria retweetledi
Geodesic is a new AI safety org i’m particularly excited about: they do awesome neglected work trying to figure out how to shape alignment priors for frontier AI. People should consider applying!
Geodesic Research@GeodesResearch
Geodesic is hiring Members of Technical Staff. Come align some AIs with us! We're a Cambridge-based AI safety org. Our seminal work showed you can bake alignment priors into base models. Now, we want to make base models robust to the adversarial effects of long-horizon capabilities RL. EOI (~5 mins): tally.so/r/vG4G6A
English
Puria retweetledi
Geodesic are a talented team pointed at very interesting problems! Seems like a great place to do open alignment science
Geodesic Research@GeodesResearch
Geodesic is hiring Members of Technical Staff. Come align some AIs with us! We're a Cambridge-based AI safety org. Our seminal work showed you can bake alignment priors into base models. Now, we want to make base models robust to the adversarial effects of long-horizon capabilities RL. EOI (~5 mins): tally.so/r/vG4G6A
English
Puria retweetledi
Geodesic is hiring Members of Technical Staff. Come align some AIs with us!
We're a Cambridge-based AI safety org. Our seminal work showed you can bake alignment priors into base models.
Now, we want to make base models robust to the adversarial effects of long-horizon capabilities RL.
EOI (~5 mins): tally.so/r/vG4G6A
English
Puria retweetledi
Excited to see @AnthropicAI landing on improving pretraining priors as a central alignment intervention. In line with our findings in Alignment Pretraining, they find that a small amount of positive AI discourse in pretraining substantially reduces misalignment.
Anthropic@AnthropicAI
New Anthropic research: Teaching Claude why. Last year we reported that, under certain experimental conditions, Claude 4 would blackmail users. Since then, we’ve completely eliminated this behavior. How?
English
Puria retweetledi
had to illustrate a situation I’ve been encountering a lot
English
It's really loud when they do that right in your face. Still had a blast filming around Cambridge with @BeestonMedia !!
Geodesic Research@GeodesResearch
Excited to see our work featured in Isambard-AI's case study series; kind of compute-intensive research Geodesic does wouldn't happen without infrastructure like this. Thanks @BeestonMedia and BriCS for the chance to talk about our research!
English
Puria retweetledi
everyone asking about his diet and lifestyle is missing the point.
every day, for the last eighty years or so, he’s woken up with a mission he’s irrationally excited about.
Netflix UK & Ireland@NetflixUK
100 years old and still the coolest person alive. Happy birthday, Sir David!
English
Puria retweetledi
Everyone loves this tweet, but it got it completely wrong. It is the sci-fi author — not the tech company — who is the true villain, for having put the story of the Torment Nexus into the training data.
Anthropic@AnthropicAI
We started by investigating why Claude chose to blackmail. We believe the original source of the behavior was internet text that portrays AI as evil and interested in self-preservation. Our post-training at the time wasn’t making it worse—but it also wasn’t making it better.
English
Puria retweetledi
Puria retweetledi
Yes! my solo-authored paper Reward Hacking Benchmark was accepted to ICML :)))
We put LLM agents in a tool-rich sandbox, give them multi-step workflows, and measure when they solve the intended task vs take unexpected shortcuts (like monkeypatching files at runtime!)
1/3
English
Puria retweetledi
