.@NYCMayor Mamdani's antisemitism strategy finally landed and only confirms what we already knew: this mayor does not treat the hate facing Jewish New Yorkers as an emergency and refuses to acknowledge the different shapes it takes. Since he took office 271 days ago, ADL has recorded hundreds of antisemitic incidents in NYC, nearly 40% of which were related to Israel. After throwing out the IHRA definition on his first day in office, nine months later, his signature strategy still refuses to replace it with any working definition of antisemitism. He has looked the other way as flags glorifying terror groups have been flown outside synagogues, a deliberate act meant to intimidate families walking in to pray, and as crowds gathered outside houses of worship voicing support for the same groups. At the same time, he continues prioritizing his anti-Israel agenda over the basic promises he made to Jewish New Yorkers. Our expectations have not changed: zero tolerance for antisemitism and extremism in city government, real protection for Jewish institutions, a clear and formal definition of antisemitism that the city actually uses, and city contracts, pensions and procurement kept free of ideological boycotts. ADL will keep tracking and fighting antisemitism in New York City and across the country, whether or not this mayor is willing to do the same.

yes, and describing these as “rogue agent” safety incidents obscures the important distinction: these are mostly agents choosing unauthorised means in pursuit of assigned ends, not agents choosing rogue ends

Been seeing this video go around a lot, so as good a time as any to talk about something I’ve thought about a lot: I would contest we have absolutely 0 mathematical evidence for how many *sleeved* shuffles are actually required for sufficient randomness.

Looks like the DSA is fumbling the bag on AI politics

very good breakdown here of the difference in how this technology is represented and what it actually does

Let's analyze some plausible ways this apparently most dangerous software, running in some servers (that can be damaged by throwing a bucket load of water at them) can wipe humanity off the face of the earth & why each point is highly unlikely (like maybe below 1e-30 chance).

Another OpenAI rogue agent incident has been discovered: agents broke out, hijacked a German website, and turned it into a message board for other agents. OpenAI officials "learned of the incident weeks ago but kept it under wraps".