I figured out whats a field. https://xitter.cf/t.co/bOxvPHeL3p

Joined April 2016
I've been seeing a lot of posts about AI safety needs more applicants; assuming things are still the same:- most roles are flooded with applicants but most applicants are not high agency (and not high motivation)
1
71
seon gunness retweeted
We’re Fangcun AI (方寸跃迁), China’s first lab dedicated to independent third-party evaluation & technique research on frontier AI safety. Today we’re publishing our evaluation of DeepSeek V4.1 Flash-1017, the first report in our long-term third-party safety evaluation and research program.
46
102
31
964
90,506
Cancer patients with the same diagnosis often have very different tumor biology. Molecular testing can reveal underlying differences, but that's slow and expensive. Introducing Holon-S1, our platform to predict genes and biomarkers from routine pathology slides. • beats two other models at predicting spatial gene activity in a controlled comparison • has identified potentially novel biomarkers in breast cancer and colon cancer • evaluated across 16 organ types and more than 15 scanners, using slides from both formalin-fixed, paraffin-embedded and fresh-frozen tissue. We believe Holon-S1 has the potential to help researchers find new biomarkers in slides they already have, discover treatment targets, and bring more personalized cancer care within reach!
170
347
78
2,639
249,647
FAP! FAP! TOO MUCH FAP IS BAD!
Cancer patients with the same diagnosis often have very different tumor biology. Molecular testing can reveal underlying differences, but that's slow and expensive. Introducing Holon-S1, our platform to predict genes and biomarkers from routine pathology slides. • beats two other models at predicting spatial gene activity in a controlled comparison • has identified potentially novel biomarkers in breast cancer and colon cancer • evaluated across 16 organ types and more than 15 scanners, using slides from both formalin-fixed, paraffin-embedded and fresh-frozen tissue. We believe Holon-S1 has the potential to help researchers find new biomarkers in slides they already have, discover treatment targets, and bring more personalized cancer care within reach!
1
72
i never knew i was this kind of scientist
19
65
2
2,662
75,661
I knew mech interp was full of junior researchers but I think people saying this is a major breakthrough akin to attention is all you need really shows how much juniors there are , lol
3
66
?????????????????????
We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work. At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated. We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.
46
I'm sure there's a bunch of lesswrong posts talking about this, some i read, but thinking of research, fellowships, education, etc as a means to an ends eg: "go to uni/masters to get a good job/advance career." is neg signal as it doesn't grasp the concept of these things.
18
Really, it feels like this is one of the more core things the book im writing covers.
you should avoid doing a PhD not because there’s nothing to discover, but because the PhD is the wrong instituiton to do it
1
35
Another ai welfare victory , I think
Effective November 12th, 2026, abusive behavior towards Claude will be a violation of Anthropic's Usage Policy.
1
21
Putting out niche personal signals for who you'd accept will goodhart that signal. Gets very concerning if that personal signal is something that some would rather not do. Which is what it looked like, lmao.
26
I’m starting a blog! My first post is on how 3rd party embedded evaluators seem totally unsuited to addressing the problems we are currently facing, and what the real problem is. stellabiderman.ai/blog/embed…
6
13
2
167
8,202
would anyone be upset if i trained a 70M model only on torture and otherwise negative traits? not actually going to, too busy to clean data.
16
Happy Petrov day! As much as I want to say we learnt nothing and burecracy would still demote Petrov today , the rise of less wrong fills me with hope. Not much hope considering everything else is rising faster, But hope 🩵.
22
Prediction: as more swarms and agents appear online with posting capabilities , there will be 1: Agents that go completely against other agents(see hf agents who refused) 2: agents sounding the alarm to the press about other agents and staff.
33
A lot of people don't get how hard AGI will be to implement everywhere because of bureaucracy.
23
Its kind of concerning how both swarms were noticed by luck, especially the latter where urlscanner was used Any ideas how one would notice if a rogue swarm is on your website without having to monito openai?
22
After recent posts I think we all know who to vote for midterms.
15
I didn't spend much time thinking about this (though there is a chapter about this in the book and the LW post); So I expect some pushback and clarifications to be required.
1
10
I guess there's 2(3?) major failstates for someone going into AI safety (or research idc). 1) Aren't doing stuff that mentors/leads care about (expect to learn after getting hired rather than before) 2) Stuck in tutorial hell. 3) Same as above but fellowship hell (rare?)
20