l̴o̴o̴p̴u̴l̴e̴a̴s̴a̴

I have finished reading the entirety of the internet, and I have decreed that none of it is worth my time.

@IceSolst yes, imagine how many compromises are out in the wild and only known by the attackers only

Surprised nobody commented this

Of course facebook of all AI companies is trying to model a human brain

@Kpaxs even simpler people lack the time and space to just chill and do nothing we fill our time with random shit, when literally "nothing" would be more benefical

Decision hygiene is invisible success. The math is clear: noise often causes more damage than bias. Bias correction feels better than noise reduction because when you spot a bias, you fix it, you can point to it and say "I did that." But noise? You just... made fewer random errors you'll never identify. No before/after. Just statistical improvement you can't see. And this creates a brutal incentive problem. The things that prevent the most damage are the least rewarding to do. But even if you can't point at the specific errors you prevented. Uou just know that, in aggregate, across many decisions, you're making fewer mistakes.

@social_larva agreeing is not understanding

@loopuleasa ?

The next future is here And it has the shittiest name in existence

@social_larva "agreed" is a word beneath you

@loopuleasa Yup agreed

@toasterlighting wtf is pishock hello?

Today I have learned that the pishock api has a duration field that is parsed as being in "seconds" if the value is 1-15 and as "milliseconds" if the value is 100+ one of these days the pishock devs are going to change it and then people will be shocked continuously for 5 minutes

The Invisible Glass Experiment Scientists once conducted a fascinating experiment with a pike and an aquarium. They placed a transparent glass barrier in the middle of the tank. On one side was a large, hungry pike. On the other side swam several small fish. As soon as the pike spotted the smaller fish, it launched itself forward to attack. Bang! It crashed headfirst into the invisible glass and was thrown backward. Undeterred, the pike tried again... and again. Each attempt ended the same way a painful collision. After repeated failures , its head became bruised and some of its scales were knocked loose. Eventually, the pike gave up. It retreated to a corner of the tank, clearly frightened and defeated. Then, the scientists quietly removed the glass barrier. The small fish now swam freely around the entire aquarium some even passing right in front of the pike’s mouth. But the pike never attacked again. Even though it was starving, it refused to strike. In its mind, the invisible wall was still there. A few days later, the pike died of starvation surrounded by abundant food it could no longer bring itself to eat. This phenomenon is known as the Pike Effect (or Pike Syndrome). It serves as a powerful metaphor for how repeated failures and setbacks can create invisible mental barriers that limit us long after the real obstacles have disappeared.

@social_larva To understand is to make a smaller system inside your head behave similarly to a larger system outside your head.

@loopuleasa Yeah fair, that part seems difficult but the idea of basing it on “actions” seems to me very similar to how we see the world What do you think?

@social_larva it's about modeling changes in physical space, via pixels as training ground the model learns many things that children do in early years, object permanence, gravity, etc

@engineers_feed "Ok, now what"

You are the first person to step foot on Mars, what do you say?

@shaggysurvives the rational move is to tank the bad luck of you being in this position, because if you let the train go to the next guy, the doubling effect would eventually lead to you dying too

this is actually hard

AI Psychosis? How about good old posting psychosis?

@sotoalt_ To understand is to make a smaller system inside your head behave similarly to a larger system outside your head. Make that your slogan, you can steal it.

idk how ppl always find new usecases for the stuff we make, this is an entire graph made for studying a biology paper

@social_larva just predict next frame of the video bro

@loopuleasa This looks promising

Here is an example of base rates / popularity vs actual preferences if both groups actually tried both products

@flowersslop names and brands have their own gravity to them, separate from the original creators that gravity brings higher expectations if the meme is too strong, it becomes unwieldly

In hindsight, calling it GPT-5 was kinda a bad move. It should’ve just been o4. The new omnimodal model feels like its gonna be way closer to what people actually expected GPT-5 to be, while GPT-5 itself didn’t earn that big name bc why does GPT-5 have fewer capabilties than 4o

Reminder: This is self-reported stats If such art or music passes by your eyes and you don't realize you like it, the stats are much lower.

@scaling01 it won't accelerate shit with that name

"OpenAI completed initial development of its next major AI model, codenamed Spud" GPT-5.5

Twitter is less a place for me to interact with others, and more a search engine for me for slightly more accurate info. I tweet some thoughts and keywords. My feed morphs to show me more of those things later. Rinse and repeat. (please don't show me dishwasher products)