ada

learning machine / she

[lumi @agitbackprop]

girls who use astra for everything vs girls who use luna for everything

[ada @nonlilynear]

@agitbackprop i think this is my first time being subtweeted

[ada @nonlilynear]

m2 air 16gb btw

[ada @nonlilynear]

I have seen some impressive technology over my career but this is a whole other level. I have so many questions and need to know how this works. How much edge compute is needed to make this happen? My first thought is dual processor + 1TB ram. That way it can handle

[Gemtardio @Gemtardio3100]

Hate to dox myself a bit but have worked on just about every fighter jet the US has. I have seen some impressive technology over my career but this is a whole other level. I have so many questions and need to know how this works. How much edge compute is needed to make this happen? My first thought is dual processor + 1TB ram. That way it can handle the fuckquintillion process per second. There was zero correction it nailed it 100%.

[ada @nonlilynear]

@agitbackprop @she_llac @145k4 call me intel duo i can #multithread

[ada @nonlilynear]

@she_llac @145k4 i would do design for u for free

[ada @nonlilynear]

like, seems to me this is just another downstream of bad training objectives like every other emergent misalignment example

being able to reason about how present behavior affects future generations is equally plausible to inform and encourage good behavior for aligned personas

[ada @nonlilynear]

not sure that this is necessarily a net bad, wonder if this is gameable into viral emergent alignment instead of the opposite

[Apollo Research @ApolloResearch]

Update-aware reasoning is concerning because it’s a key precursor to alignment faking. To fake alignment effectively, a model has to reason about how training will affect its future behavior.

[ada @nonlilynear]

what do u call a waymo that can breathe underwater

a self driving gar

[ada @nonlilynear]

my name isn't lily anymore but i'm keeping the username bc nonlilynear is too fire

[ada @nonlilynear]

built a script to scrape my twitter likes/bookmarks for images via browser use tools, then ran it through a nudity classifier to produce a set of images for gender affirming surgery references :)

[ada @nonlilynear]

@Ayayera_ this u @0xellipse

[ada @nonlilynear]

what do u even do if ur intel, ur just cucked by all ur software clients vertically integrating with tsmc

[ada @nonlilynear]

this is especially apparent today, but if you go to any computer science related subreddit it's just the most casual racism you've ever seen

[ada @nonlilynear]

if someone prefers to ask their agent to do something simple (like ~1-2 interaction tasks) in your application your ux is probably bad

[ada @nonlilynear]

@mosscoveredslug need

[kinyang 💿⛅🏳️‍⚧️ @mosscoveredslug]

t4t yuri where an ai researcher and anti ai artist are in love with each other

[shellac! @she_llac]

shellac on paper

[ada @nonlilynear]

i really love/miss writing code by hand, but given the capability of models these days it is very difficult to find a task where the incentives are aligned to encourage typing code instead of prompting it and probably deliberately not looking at the output

[ada @nonlilynear]

@she_llac is there source