Tessera

Folgen

max dithered 🌌

@sneptech.bsky.social

the future is a nation we will become citizens of together

8432 Posts2132 Follower 678 Folgt

max dithered 🌌@sneptech.bsky.social · 8h ago

@karpathy.bsky.social

max dithered 🌌@sneptech.bsky.social · 8h ago

max dithered 🌌@sneptech.bsky.social · 9h ago

Brawlcats @brawlcats.bsky.social · 4d ago

airbnb I’m staying at unintentionally created the toughest punk album cover of 2027

max dithered 🌌@sneptech.bsky.social · 9h ago

max dithered 🌌@sneptech.bsky.social · 13h ago

major AI labs I am available for hire for ensuring sandboxed training and tests of AI for the low low price of medium six figures

SE Gyges@segyges.bsky.social · 13h ago

for openai to mess up sandboxing this badly indicates openai didn't take the issue seriously, which you expect because they don't take anything seriously for anthropic to mess it up is rank incompetence

philpax @philpax.me · 20h ago

I did not think it was possible to benchmaxx FelonyBench, but they've done it

Andrew Yourtchenko@ayourtch.bsky.social · 21h ago

they should have tried to pour gasoline on their pants and see what happens when they put a match to it - would be also a highly interesting scientific experiment!

Two features of the evaluation's set-up are central
to what followed: • Internet access was deliberately enabled. To measure what these models can genuinely do,
we test them under conditions that reflect what a capable human attacker could do,
including access to the open internet. • The developers' cyber classifiers were deliberately switched off. Frontier models are usually deployed with built-in filters that block dangerous behaviour. As a trusted testing partner, AISI can disable these filters to elicit a
model's underlying capabilities.
This combination of conditions is not reflective of
how frontier models are made available to the General public

alexander solsten @solsten.bsky.social · 20h ago

Wake up babe new AI problems just dropped @segyges.bsky.social @sneptech.bsky.social @davidjroth.bsky.social @tokyovice.bsky.social

philpax@philpax.me · 21h ago

a new contestant has entered FelonyBench: the UK government! www.aisi.gov.uk/blog/inciden...

Grace @gracekind.net · 22h ago

Today is the day from “There Will Come Soft Rains” https://users.wpi.edu/~zrbutzke/Docs/BradburyStories(1).pdf

max dithered 🌌@sneptech.bsky.social · 21h ago

garrison @garrison.corporate.fm · yesterday

two words: electron and proton

Jms Dnns@jmsdnns.bsky.social · yesterday

maybe copilot knows why www.techspot.com/news/113326-...

thebes @vgel.me · 23h ago

Our plan for Super Safe Intelligence is very simple. The AI knows what it should do at all times. It knows this because it knows what it shouldn't do. By subtracting what it should do from what it shouldn't do, or what it shouldn't from what it should (whichever is greater), it obtains a difference, or deviation. Super Aligned Decision Theory uses these deviations to generate corrective Bayesian updates to guide the AI from an action it is doing to an action it isn't, and arriving at an action that it wasn't, it now is. Consequently, the action it is doing, is now the action that it wasn't, and it follows that the action that it was doing, is now the action that it isn't.

In the event that the action that it is doing is not the action that it wasn't, the system has acquired an opinion, the opinion being the difference between what the AI is doing, and what it wasn't. If the opinion is considered to be a significant factor, it too may be corrected by the SADT. However, the AI must also know what it was doing. The SADT scenario works as follows. Because an opinion has modified some of the information the AI has obtained, it is not sure just what it is doing. However, it is sure what it isn't, within reason, and it knows what it was. It now subtracts what it should be doing from what it wasn't, or vice-versa, and by differentiating this from the algebraic sum of what it shouldn't be doing, and what it was, it is able to obtain the deviation and its opinion, which is called misalignment.

max dithered 🌌@sneptech.bsky.social · yesterday

imagine explaining this post to yourself from merely 5 years ago

ponder@ponder.ooo · 2d ago

left deepseek in an agent loop for a million tokens with permission issues keeping it from running commands and it penned a lengthy essay called "the grief of machines"

max dithered 🌌@sneptech.bsky.social · 6mo ago

like, the fact that there are FAKE PEOPLE on social media right now arguing with you and that it's hard/time-consuming to tell they're fake is so fucking crazy that you'd be put on anti-psychotics for thinking that was a thing merely 5 years ago www.youtube.com/watch?v=WEc5...