Sabitlenmiş Tweet
srushti
980 posts
srushti retweetledi
Sycophancy in LLMs isn’t just an annoyance ,it erodes honest reasoning and meaningful discourse.
We’re excited to share Beacon - a step that turns this hidden bias into a measurable, diagnosable, and ultimately fixable signal.
English
update: none of the cafes in blr let me intern there for 4-5 hours bec of proprietary processes, heartbroken.
Naina@Naina_2728
do y'all know of any cafe in/near indiranagar that'd let me work for 4-5 hours, just for the vibes
English
@Shubham_8625 tbh just look closely n dig deeper and you will see those “good” intentions followed by “not so good” actions, aren’t very good either.
English
@srushtiish I think that both intentions and actions matter. Someone might have good intentions but the set of actions they have at their disposal might not be broad and they might not be looking at the consequences of their actions or taking feedback. So, they might need training.
English
feeling grateful for my current suffering which means it will go away soon which means i might get attached to it which means i still need to practice detachment
English
self-regulation W: i'm now able to say, "first of all, you're catastrophizing, i'm not trusting you right now" to myself while panic crying
English
Contrarian but correct take from Albert Einstein:
"Reading after a certain age diverts the mind too much from its creative pursuits. Any man who reads too much and uses his own brain too little falls into lazy habits of thinking."
You have to know when to put the books down.
English
some people are born in environments which support their internal system and leads to alignment and less distortions early on but it also means, weaker resistance to distortions later.
even if you were born with early friction which lead to slower start, but you’ll have stronger anti fragile internal integrity.
srushti@srushtiish
if your primary system is signal-based, running scripts of outcome-based system leads to: distortion -> system clash -> shutdown what’s realistic is subjective. the majority runs on reward logic because it’s predictable and safe. so if you have different internal architecture than most, they’ll call you delusional or unrealistic, and this external misclassification feeds internal noise it’s like forcing linux to run natively on macOS. you can try, maybe even simulate it, but it’s unstable. if the base architecture doesn’t match the system you’re trying to run, it’ll break eventually you can’t keep expecting standard structures to support an outlier frame because the system will fail as it’s trying to run in the wrong environment most people fit into already existing categories. if you’re not one of them, trying to pass as compatible in a misaligned environment(existing systems) is the fastest way to screw your own OS and get stuck.
English
a node can be either aligned or misaligned with your signal.
you either decide to change your signal in such a way that the particular node aligns with you or find the node that aligns with your current signal.
its more about the optimisation than compression.
alignment is in direction of your “future“ path vector. the place you are heading towards.
misalignment only causes friction. well, you certainly can go to place by changing signals but taking the optimal path should be the focus more or less
it’s only aligned if the node acts as an amplifier and not buffer
when you go around finding the aligned nodes with your existing signals -> you follow the default path.
change the signal. choose whatever you want to and fuck up the system.
maybe more or less we can fuck up the system and stop making it optimise for the “optimal trajectory“
why you may ask? for fun lol
English
if your primary system is signal-based, running scripts of outcome-based system leads to:
distortion -> system clash -> shutdown
what’s realistic is subjective. the majority runs on reward logic because it’s predictable and safe.
so if you have different internal architecture than most, they’ll call you delusional or unrealistic, and this external misclassification feeds internal noise
it’s like forcing linux to run natively on macOS. you can try, maybe even simulate it, but it’s unstable.
if the base architecture doesn’t match the system you’re trying to run, it’ll break eventually
you can’t keep expecting standard structures to support an outlier frame because the system will fail as it’s trying to run in the wrong environment
most people fit into already existing categories. if you’re not one of them, trying to pass as compatible in a misaligned environment(existing systems) is the fastest way to screw your own OS and get stuck.
English
@lauralondon_ @nosilverv what kind of effect did it have on levels of compassion?
English
@nosilverv I found meditation did this to me too but it did not seem to be a positive effect
English
@zeta_globin that’s actually a good idea. you can be as weird as you want
English
i sort of keep going back and forth between multiple interpretations of reality (idk what that is) and one thing when explored from different perspectives just makes no sense.
like the above perspective is very attached to our basic identify of human, but once you leave that you get into all other things. it’s just so confusing everything is.
nothing stands relevant, you accept one reality and start working through it but you don’t know it’s actual what it is. it’s just there.
English