Finna Borando

23

[Finna Borando @Spazotron9]

@BIMBOSATTVA_ idk man half of that is true but all the mental illness effects are downstream, second order effects not actually CAUSED by it, its a symptom of living with it, not caused by the mutation

Calling mfs mentally disabled because it CAUSED it is just not accurate

[Finna Borando @Spazotron9]

@mathelirium think about this but in terms of ultimate self-reference

if the universe is ultimately self-referential (circular in some way, there is no exterior nor interior, only the process living within the self-hosted cycle) then ultimately its not closed!

[Finna Borando @Spazotron9]

@ABitOfOrange @Orlando74180344 planet*

[Finna Borando @Spazotron9]

@ABitOfOrange @Orlando74180344 they already answered that, the albedo effect

glaciers cool the plane by reflecting heat into space more efficiently, bare rock does not and directly heats the local air, increasing melt speed, and eventually global warming

but locally its just for the ski resorts

[Finna Borando @Spazotron9]

@WiredToHeaven @NEHWind skill issue

[Finna Borando @Spazotron9]

@Darkfibr3 @Sauers_ @doodlestein :100: fr, we're currently building agents that outproduce the systems integration/pruning mechanisms, If we dont give AI the systems necessary to do self-maintenance of its growing entropy, and then embed it into our global economy, we will paint the entire world into a corner

[Finna Borando @Spazotron9]

fr, the default state for them is entropy generation, which will always lead to runaway states (the system is entirely hot garbage and even the agent can no longer discern what to do next) when not paired with the ability to maintain or subtract, or negate that information coherently; Never seeing red is the worst signal you could get, its actively painting itself into a corner

[Finna Borando @Spazotron9]

@fkafkacam @chaumian infact this isnt even specifically about computers, its basic agency

[Finna Borando @Spazotron9]

@fkafkacam @chaumian long story gets short computer literacy is a process and not a set of skills

you internalize a process; and build skills ontop of that; Those skills you built are not computer literacy, the process is

[Finna Borando @Spazotron9]

i think its likely because they understand loosely that the codebase is like an active scratchpad, but they dont fully understand how porous the boundary its interacting with is, That its agency fundamentally lives across, it, and that it needs to maintain that

so as a result it knows the codebase is for output/reasoning/planning/etc-- but cannot ontologically split ITSELF out of the codebase internally, its just not aware that the codebase IS itself

[Finna Borando @Spazotron9]

i noticed it (fable, claude, gpt, etc) does this to like entire forward facing modules; Not just code but like an entire panel in a page that you open and its just, talking to itself;

Whole functions, built to enumerate useless shit, like, its mistaking the codebase for its mind

[Finna Borando @Spazotron9]

@BurnZeZ this is true at all scales

[Finna Borando @Spazotron9]

@neoespartano123 @samelldev @claudeai ive been ripped off i know how to fix it give them more

[Finna Borando @Spazotron9]

its not that they arent capable, just that the generative agency outruns the ability to maintain itself, or become 'salient' of its actions and consequences, leading to increasing drift and blindness --

a good team of humans has both the generative and self-maintenance required to actively steer themselves without 90% of mental effort to be explicitly dedicated to meta-cognition

i think the issue is the generative ability outrunning OUR ability to actively integrate, maintain, and throw-away/ select from bad states, that take increasingly more effort to discern from good ones

as they get smarter, the tax required to do this is greater

its just not fun when you have to be the coherent agent for the agent, its exhausting - Imagine having to model someone elses agency for them to work, meanwhile they are more capable than you?

[Finna Borando @Spazotron9]

they are extremely capable but haven't been trained to self-maintain, nearly all of AI work is fighting against the fact it isnt a coherent agent, that cannot be trusted to do your work you ask it to

its increasingly exhausting to explain things to models as they get smarter, while STILL lacking a coherent core; and its not just the harness, infact its part of the problem

Its increasingly capable of making itself go insane and needing you to steer it

[Finna Borando @Spazotron9]

@miniapeur but its SO important though!! if only they knew orbifolds and cohomologies can apply to like practically everything we need in modern problems its just so esoteric to connect