I just figured out why Anthropic decided it could change their RSP instead of stopping like we all thought they promised - it's because they released the updated RSP on April Fools, last year, and it turns out they were always just joking about it being a real commitment.
We’ve made some incremental updates to our Responsible Scaling Policy - these updates clarify our ASL-4 capability thresholds for CBRN, as well as our ASL-4/5 thresholds for AI R&D. More details here: https://t.co/3b4Kgp0s1U
Build your own control panel for reality.
Many sensors, varied visualizations, multiple dashboards, all focused on monitoring your ship staying on a single coherent course.
Irony is a sandbox for sincerity.
You can speak ironically in private until you’re done fucking around. Hopefully once you finish, you can say something real in public and actually mean it.
The goal isn’t a single worldview that fits every situation.
It’s the freedom to swap ontologies mid-problem when the old one stops being adaptive - and wisdom is the ability to see when that occurs.
Make your beliefs true enough that they withstand honest attempts at falsification, yet beautiful enough that you cherish them even when others reasonably disagree.
Be sincere in public.
Irony is for mentally stormy days stuck at home with friends, not as a broadly shared epistemic norm in rapidly changing and varied climates across disparate groups.
Question beautiful widely accepted purported wisdom; wise sayings persist not because they are true, but because beauty is easy to remember.
Instead, embrace heuristics you see others use successfully, strive for boring but falsifiable insights, and build adaptive plans.
When truth gets slippery, find the part of you that listens, not the part that argues.
Then as long as it is working, scale that up to internal cooperation, external outreach, and if you're lucky, useful insights to share with others.