天上的锦鲤

@xenocosmography well it was about time that the information age would actually bring important information closer to the ones in power we do not need that gap of decades of information anymore

Have to confess this is a little weird. x.com/matthewjmandel…

@louszbd do you feel OpenAI vs Anthropic is similar to, respectively, Microsoft vs Apple? in their approaches

views on my own: 1. OpenAI and Anthropic are diverging. OpenAI is going full vertical integration: from custom chips to healthcare, social, owned infra. Anthropic: no custom silicon, infra outsourced, laser focused on making the best model and API. Claude Code getting all the love. both are valid strategies, will be fascinating to see which wins:) 2. need (more) "AI success engineers" on the ground helping people deploy. 3. OpenAI has hardware + ops roles in Singapore. the hardware supply chain is quietly extending into Southeast Asia. 4. xAI has 27 in-house annotation roles. everyone else either outsources or keeps it quiet. the team might value data quality control enough to eat the cost overhead. 5. both Anthropic (Universes) and OpenAI (Synthetic RL) are building teams around ultra-realistic long-horizon environments for agent training. the quality of RL training environment might end up being the deciding factor in whose agents work. epoch.ai/gradient-updat…

@Ablazeolive me when I can't read a master thesis

Why does everyone trust this WPA stat so much? This is obviously nonsense lmfao

@Ablazeolive @zac_gaming__ imagine not understanding what WPA means

@zac_gaming__ People are gonna see that post and think "hm I should try sapphire crystal start" Nothing wrong with trying new things but it is clear to any good player that dring is way better. WPA is incorrectly indicating which item would be better to purchase

"why does everyone trust WPA? 🤓" please just read xPetu's thesis and his videos, stop being this dense. if you can't even try to logically explain your reasoning, why do you even speak?

stop being funny and fix it dude

🇧🇷🚨 Documentos da WikiLeaks revelam que Othon Pinheiro, o maior engenheiro nuclear brasileiro do século XXI foi apelidado de “czar nuclear” pelos diplomatas americanos, e tentou criar um mecanismo paralelo de inspeção para driblar as exigências intrusivas da AIEA. O objetivo? Proteger os segredos do programa nuclear brasileiro.

Ngm acha razoável mandar negros voltarem pra senzala, mas desumanizar mulher é quase um direito pra maioria das pessoas.

@tak_01_ *an advantage

@tak_01_ guarantee and advantage in the middle-game and the actual 'play it out' stage.

thinking to myself does meta even exist if you have a cohesive comp that counters enemy team esp in individual lanes does it even matter if you picked lee sin top and taric support if it works i dont think there will ever be a champion that is worthy of being permanently banned or picked purely from the opportunity cost of doing so so in my head why still have default comps just take 2nd pick and beat whats in front of you

@tak_01_ ( you can play a 'bad gambit' in draft, get punished by the other team [even if unintended and not planned out] and still win, so, why change? )

@tak_01_ not comparable in practice since it's not rewarded at all, but I hope my point comes across

Slavery is bad? Not for everyone. Look at the results of the UN GA voting on the relevant Resolution. See which countries don't support it and ask yourself why? Colonial way of thinking is still in their minds. #400years #Slavery #freedom

Godard’s classroom in La Chinoise about Two kinds of Communism

March 27th: your chance to see yours truly on the big screen 🫡 Movie ended up with good balance of optimism in the end!

@Ambisphaeric (and this implies superintelligence would be *able* to be confined to a 'ruler')

The harm that could come from immoral or highly opinionated people (who may disagree with you) governing these technologies can't reasonably be overstated. I think even one person with a bit of agency could already be far more dangerous than one might expect. This is a huge risk area that needs to be addressed, but I don’t really want to go into the specifics unless someone wants to open that door. Most simply aren’t ready to hear it. There’s a very good reason (IMO) why SOTA models from the labs aren’t released with open weights or are put on some kind of delay. Even then, what’s already out there is incredibly powerful as-is. This is because the technology, while not inherently dangerous, becomes risky in the open world precisely because of the capabilities it gives to users who have agency. The governance of superintelligence therefore needs to be handled by moral people whose primary role is ensuring that no immoral actors can weaponize a GPU to harm another person or group. That would require either “someone” trustworthy to hold the keys to the most advanced superintelligence - or for everyone to share access to it, in both the offensive and defensive sense.