AISI got an especial carve-out for those technical roles, which is why they are somewhat competitive, but the pure policy roles are not as high. I can’t find any current AISI policy job openings but in the past a Transformative AI Policy Analyst had a salary range of £44,195–£48,620, and a Strategy Delivery Manager, a range of £44-48k
Rasool
Karma: 135
[Question] Where do you point newcomers who want to get involved in AI safety?
Further confirmation about 5.1/5.0/4.1/4o etc using the original pretrain from this Time interview with Greg Brockman
At 8:20 (emphasis mine):
For a period of about two years, right from when we announced GPT-4o in April of 2024 to December of 2025, we were really squeezing the juice out of the GPT-4o model, right? that that we sort of had a bunch of pieces that uh we were improving but that fundamentally we were on the same underlying base model
I’m surprised to hear you describe the effects as modest in poor countries, GiveWell are fairly positive on them, where does your disagreement come from?
I am a happy user of espanso for this
Nvidia is hostile to Anthropic in various ways, despite investments.
What is this in reference to?
The 2028 Global Intelligence Crisis—a finance-oriented vignette
This person has been keeping track of their dealing with the BBC wrt to their choice not to get a TV license since 2006 (I believe at some point their SEO was so good that searches for ‘bbc TV license’ brought up this person’s website, rather than any official source)
They document things like the technology claimed to be used, and the results of Freedom of Information requests
Other fun pages include bogus/contradictory numbers supplied by the BBC, people spotting vans in the wild, and why the BBC’s threatening letters request people to not ‘write below this line’
Anthropic’s Frontier Red Team recently wrote about how Opus 4.6, fairly autonomously, “found and validated more than 500 high-severity vulnerabilities” in open-source projects
The UK Health Secretary in 2021, Matt Hancock, ordered 100m vaccines, rather than 30m, because of the film Contagion
A few days ago, Amodei claimed that 90% of code at Anthropic (and some companies they work with) is being written by AI
https://www.themidasproject.com/watchtower tracks changes to AI company policies
Great piece! One minor detail is that Llama 4 curiously only used 32k H100s, rather than the 100k+ mooted
The Midas Project is a good place to keep track of AI company policy changes. Here is their note on the Anthropic change:
https://www.themidasproject.com/watchtower/anthropic-033125
Also picked up here
Do you think there is anything to the idea that Mythos/Fable/Astra are smaller than 10T parameters, in order to increase their inference efficiency for longer horizon rollouts?
This is the view espoused for example by Charlie O’Neill and Beren Millidge on the recent Dwarkesh podcast (emphasis mine):