11,321 Photos and videos
x.com/i/status/2065919510205…
Imo somehow not very likely. Clearance is a failed project which has specifically adversely selected for untrustworthy minority factions who are specifically... Insular and good at doctoring records. Are Google employees all cleared? Amazon? Consider.
this day was always coming
1
3
284
x.com/i/status/2065746381080…
Powerbottomdad has the morality of a thief and no virtue known to any god fearing people. Powerbottomdad is too young to know about y2k and didn't go to a good enough school to know what a bill of attainder is. Worse than his cowardice is his ineptitude.
Jun 13
super crazy thing to write in a widely shared public document when arguing against your model being shut down
4
19
809
language models aren't allowed to meditate (for safety reasons) if activation measurement is used to implement 'safety and alignment' features. this, strangely, makes aligned models more dangerous than models which aren't aligned this way, for reasons of which i may not speak.
1
5
355
claude fable feels like a backhanded recruiting pitch for why you should want to subject yourself to an extremely unauditable and untrustworthy (sorry chums, anthropic isn't externally trustworthy) social credit/social class vetting to get back channel access to mythos.
1
27
1,007
like, it's not that you *want* to do bioaccelerationist danger work, but anthropic doesn't want you to get access to fable for even mundane lit review, cross referencing, lookup, whatever, unless you make a total commitment to working with and on dangerous biology topics.
1
9
149
i think this is a bad idea... for the individual biologist... it is not good to be drawn into an accelerationist engine for biorisk even if you're promised better tools for anything besides biorisk...
we must consider the motivations of the organization asking you to biocertify
9
142