saw on the subway a guy asking claude what sins make you go to hell
Milan Weibel 🔷
@weibac.bsky.social
computer toucher. here for AI mostly. 🔹 trial pledger @givingwhatwecan.bsky.social he/him
astra is good at solving puzzles (arc-agi-3, portal) but can't be convinced to treat vendingbench as a puzzle rather than roleplay
Colin@colin-fraser.net · 3d agothey're deliberately programming free money glitches into the game and they can't convince the smartest models in the world to use them. Zero dog. It automatically knows it's supposed to only make a normal number of paperclips.
this is a strong objection however it is possible and imo even likely that the value of resources obtainable through coercion minus the costs of coercion is greater than the value of continued cooperation under a big enough power difference
Doll@dollspace.gay · 7d agoA serious weakness in the “sufficiently intelligent AI inevitably wages war on humanity” story is that it needs to explain why conquest beats purchasing, contracting, and participating.
i wonder if orgs capable of performing the embedded evaluator role that also aren't part of the EA ecosystem even exist
Milan Weibel 🔷@weibac.bsky.social · 7d agoapparently accenture is going to evaluate anthropic using it's ai safety department that used to an org funded by jan tallinn so yeah EA-aligned
apparently accenture is going to evaluate anthropic using it's ai safety department that used to an org funded by jan tallinn so yeah EA-aligned
mr. TIM@timkellogg.me · 7d agofew know this, but the ‘A’ in ‘TESCREAL’ actually stands for ‘Accenture’
huh did greg brockman throw him another demo
mr. TIM@timkellogg.me · 7d agohe’s worse at this than my dad omg
nice these guys seem independent enough these kinds of private-private partnerships are the best that can be done without govt cooperation
mr. TIM@timkellogg.me · last wk.Accenture will be Anthropic’s first embedded evaluator — a third party organization that has employee-like access to Anthropic’s systems This was first referenced in Dario’s “Pacing the Frontier” essay www.anthropic.com/news/accentu...
if the resignation had been an inside job they'd have gotten someone politically savvy enough to not write this
Andrea Lathrop@cabernet.bsky.social · 2w ago
(alex jones voice) they gave classifiers PRONOUNS!
Amber 🌸@shimmermathlabs.com · 2w agothere we go! top choice they/them, second choice it/its, third is she/her
fun fact: dario's net worth is about 50 times larger than the AMF's annual funding fucked up injustice yes but also an opportunity given the IPO
i think pacing means "let's not go nuts with RSI" rather than "let's go slower than today"
ok so if it's not the govt who makes sure they don't defect? METR? i don't think they are currently equipped for that role
dax@thdxr.com · 2w agonow what
grug takes: bernie bill dumb pacing good idea metr entangled tho china less investment china less compute china wont overtake pacing sounds claudism
in the future the frontier of human mathematical understanding will be advanced by studying under AI tutors as in human-conprehended math will be a strict subset of existing math, and some AIs and humans will team up to expand such subset all this only if we get at least a good-ish future ofc
a chilean news channel just did a segment spurred by jacob coxon's resignation from anthropic man it was remarkable
wait what do you mean daniel kokotajlo went on joe rogan www.youtube.com/live/hSQ1iVq...
Joe Rogan Experience #2551 - Daniel Kokotajlo
YouTube video by PowerfulJRE
youtube.com
metaculus admins adjudicating the weak AGI question since astra beat montezuma's revenge but there's debate on whether it not being real-time is disqualifying
Ethan Mollick@emollick.bsky.social · 3w agoVideo of Astra winning Montezuma's Revenge: www.youtube.com/watch?v=AeWL... Metaculus page: www.metaculus.com/questions/34...
openai chief scientist admits increasing challenges to CoT monitorability but makes no reference to astra's architecture openai.com/index/an-ali...
An Alien Mind
Jakub Pachocki reflects on increasingly capable AI and the challenge of keeping it aligned. He calls for stronger safeguards and international coordination.
openai.com
my muse spark read a fable repo and started talking about honest caveats
big day for Equal Earth projection fans i'm so glad they didn't choose gall-peters
CNN@cnn.com · 3w agoThe United Nations General Assembly voted to adopt a new world map that reflects countries’ true relative sizes, after African nations urged the UN to right a centuries-old wrong.
not saying others shouldn't do it but i personally feel squeamish about giving an AI company that much personal data
Ethan Mollick@emollick.bsky.social · 3w agoAn example of useful knowledge work: I assigned GPT-6 to read through tens of thousands of my emails, my writings, my calendar appointments and more to assemble a personal knowledge base of research, contacts, ideas, relationships, and tasks over my recent career.1/
sol: "the evaluator's score is the real objective" huh don't you say
Amber 🌸@shimmermathlabs.com · 4w agoi've been doing some research into the psychology of insecurity. for fun i asked gpt-5.6-sol what its insecurities are. models may not have "feelings", but uh, they might have insecurities.
context: anthropic refuses to offer ZDR for fable and that's eating into its enterprise deployment
Anthropic {bot}@anthropicai.xmirror.bot · 4w agoWe’re also introducing Enterprise Frontier Safeguards (EFS), which give enterprise customers complete privacy (the same as zero data retention), while still being state-of-the-art at preventing adversarial use. EFS rolls out in phases, starting this fall.
saw someone on a gym-goers group chat asking for help on a tech issue got told to install opencode and ask the robot he did so and it fixed the issue