@realsudarshansk@ycombinator@garrytan to be fair, the average founder IS cooked, this is a 0.1%er industry. YC is not a charity. That being said, they do make mistakes, and you do not need their permission to build a unicorn
Brian, for the love of god, you cannot take a health database, click 'sort' and think that the top 5 drugs patients who survive longer happen to be taking are causal to the benefit those patients received. 500k is actually not a large cohort for a database, needs alpha correction, was very unlikely to be prespecified, etc. that's why no one is impressed with these studies and they're published in trash journals.
this study showed a slight increase in CV events with PDEs vs placebo
https://t.co/y3AbyTFHUU
the mechanistic rationale is not there and your explanation is terrible. ALL PDEs metabolize cAMP/cGMP, thats why they are phosphodiesterases. are you suggesting we should inhibit all PDEs?!
by the same logic we should all be taking ERAs too. why not ARBs and ACEs? screw it i'll take inhaled treprostinil too. might extend my life. then i'll run SQL queries on health databases until i see a 'signal', bonferroni may roll in his grave but i will be vasodilated.
NEW: malware developers added nuclear & biological weapons text to to their spyware.
Goal? To trigger LLM safety refusals... so that their spyware wouldn't be analyzed by an AI security scanner.
Cleanest practical example I can think of for why over-indexing on first order safety alignment is risky.
When closed (and open) models ship with aggressive refusals, they will be sprinkled with second-order blindspots that attackers will discover...and exploit.
We are only in the earliest days of attackers leveraging these features, and it wouldn't surprise me if users systems that need to handle complex cybersecurity issues demand that models be less safety-blunted.
In the weeds: @SocketSecurity's post also shows why intention matters in how you design a malware analysis pipeline to avoid prompt manipulation.
H/T to colleagues that shared this with me https://t.co/f3Aj9TYxU4
NEW: Anthropic is walking back Claude Fable 5's policy to covertly degrade performance for competing AI researchers, after facing fierce backlash.
“We’re changing Fable 5’s safeguards for frontier LLM development to make them visible,” Anthropic tells WIRED. “We made the wrong tradeoff and we apologize for not getting the balance right.”