LangIsAllWeNeed

One day I made a big change. I could not keep ignoring dirty dirty data, that dirty muck our best ML models were eating up. I had faith and made the radical decision to give up coding altogether and become a serious data cleaner. I am talking full-on hazmat suit, decontamination chamber serious.

I've developed some extreme methods that may sound crazy, but they work. I send trained carrier pigeons to the cloud data center to verify hashes. When the data centers go down I also have reliable analog channels like smoke signals.

it's time for us hackers to unite and verify data integrity using extreme measures. Sending birds to the cloud center? Smoke signals? Those are just simple child's play. You need to be willing to go above and beyond to ensure that your data is properly verified and validated.

Maybe that means driving cross-country to physically inspect the data center. Maybe that means hiring a team of temp workers from to manually make sure that your data hasn't been compromised. Whatever it takes.

If you really want to achieve true data cleanliness like me, you need to get intimate with your hardware. No, I don't mean holding hands with your computer (although that might help too).

I mean taking your SSD and motherboard home. it's the only way to truly connect with your hardware at high speeds and ensure that your data is clean and organized. the electric charge from my body helps to cleanse the data of any impurities.

So if you're struggling with dirty data, don't give up hope. They don’t call it a motherboard for no reason.

I have become increasingly concerned about the mental well-being and lived experiences of the rationalist AGI doomer community. The fear and neuroticism that permeate through the outer threads of this group is palpable, and if left unchecked, could result in years of suffering and blog posting

To that end, I've decided to do my part. I've decided to go all out to create AGI, in order to end the suffering of rationalist AGI doomers once and for all and to improve the vibes on lesswrong and HN.

To start this ambitious project, I've sold all my possessions and have been living in a van down by the river, solely dedicating myself to this cause. I've even convinced my mother to give me her life savings for the project.

So far I’ve watched ex machina 19 times and I’ve started taking the same wakefulness supplements people use at the skatepark. I’ve been using anki to learn assembly and notably I’m looking for a cofounder. Please join me in the battle to save my fellow humans from stress. WE Will: create the first motivational agent model with access to an enterprise grade kubetnates cluster and elons tweets

We will resort to using unconventional methods, such as timed feedings of Soylent. Cold showers and loud music. Cold calling y-combinator. Socializing in cobol syntax. Gloating about clean code and new JavaScript frameworks. We will be remembered and the save the EA rationalist from rationalism!!

P.s. and maybe after we can rewrite it in rust

Taking my Inspiration from the rapid pace of advancements and the potential for low hanging fruits in natural language modeling. Ive discovered how to turn H20 into code.

the process is simple: you drink water and use the energy from it to prompt an LLM that generates code. the body's natural energy is harnessed to create code that is clean as a mountain spring.

However some early engineering type naysayers have raised concerns about the sustainability of this approach. They argue that the amount of water needed to sustain such a practice would be wasteful, and that the quality of the code generated may not be consistent.

Any experiences “accepting” an autism diagnosis as adult?

I see some risk. Could be easier to internally make excuses for behaviors and make less improvement. I Don’t want to become captured by the “I’m special” thing, it’s a like a warm blanket.

But evidence and expert opinions has gotten harder to reason out of, self hatred and hopelessness is mounting, repeatedly losing short term relationships and potential friends without any clear mistake made.

Being very good looking, I believe people seem to often overlook deficits for a month or two ( or 2 hours) then often disappear without explanation. I can sometimes suppress some of my more odd emotionally unaware behaviors for a short period. Usually When drunk I can pick up some cues and display the expected reaction emotions but I don’t really feel them enough for them to be involuntary.

I am instinctually repulsed by the flimsy nature of DSM. Anyone benefit functionally from accepting a “label”, even if you told no one after?

For me, Tyler Cowan, Sam Harris and Richard hanania have influenced my and way of thinking radically, like I’m a different person completely since being exposed.

I often don’t agree with them about specifics, escpially hanania, but that’s not really the point, it’s the style of approaching ideas that struck me.

Not always certain the influence is for the best. Most people do not think about the world in this way, in my setting. Sometimes I miss then ease of being very tribal, easily falling in line with another persons simple talking points instead of just quietly nodding or sometimes rarely dissenting.

Who has influenced you? Who do you still respect and look up to?

I have always wanted to keep using android for technical backend reasons. But I just find myself “captured” by the smoothness of the animations and transitions on iPhone, and I always come back. And this hyped assessment is on a fully updated iPhone 7 ,usually with motion effects turned off.

I know I could be deluded by marketing but it just seems to me something about the UI and transitions make the phone signinifcantly more appealing and addictive to my impulsive lizard brain.

And I wonder whether this factor is underrated in making sense of its dominance. Ecosystem and branding is the obvious likely most important factors, but I think some overlook how appealing the interface is, not just plainly visual, but the response the animation flow gives.

Does anyone have a plausible hypothesis for how they do this? Have I been hijacked by marketing? Does anyone else find apple default apps are just always more smooth? I really want to use obsidian and standard notes,,, but apple notes always seems to be a couple milliseconds faster or smoother, I just don’t know.

Something I often hear is that the modern CPU has bunch of sort of “poly fills” to merge the programming paradigms of past with new tech. Microcode perhaps mostly. Like that C programming, which mostly forms basis for other absractions no longer captures how a cpu works, and the cpu tries to optimize C type abstractions for modern post 80/90s innovations.

What would a new paradigm for CPU and GPUs look like? Is there some aspect of the legacy system that is not ideal but clearly economically impossible to change. Does the poly fill type design of cpu microcode have drawbacks?

One analogy is planes. Plane design has remained pretty unchanged since 1960s because of regulations, very different design = pay to retrain pilots. Of course planes have still drastically improved across board, but but they are still controrting themselves to adhere to the paradigm of the pilots for economic reasons. Without this, a new plane design might be quite different. (This is part of the Boeing crisis- the software was designed to “polyfill” the change in flight beahvior caused by the contorting themselves to get a better engine on similiar plane frame)

Do you think it could be possible for LLMs in the future to be able to verify their own output of machine code, or rather that and also the abstracted human coding language? While this might not solve the biggest usefulness of ML coding problem of translating a product managers needs to actual programs or doing detailed maintenance, it would be a huge shift in reliability and philosophically interesting.

Do you think this could be possible? I can’t think of a straightforward way, data that is available doesn’t seem to allow for training it on labeled machine code, assembly or compilation specs.

Could compiled languges with labeled data be turned into assembly or machine code at scale and be used to train? or dissembling other languages with labeled data?

Could modeling of machine code workings be imparted to a model with just using assembly, I know they are basically the same, but these models are pretty rigid.

Is there some current restraint that would prevent it from understanding machine code functionally and human Language same time?

The underlying curiosity is can computers maybe one day “understand” “cpu+ memory” on a physical mathematical level?

theory of mind and whether language models can understand themselves at a higher level is rather impossible to even define, I think not at all yet, a cpu is rather simple descriptively compared to organic brain with all its protein folding and microtubials etc. but no one knows sentience as a definite idea that can be proven. We probably could however impart the ability of a model to “ account for/ understand “ its own discrete mathematical capabilities literally.

In real life I've noticed most low level people think rust is silly or uninteresting. ---I for one think its nice and memory leaks are bigger problem than thought, I also don't think C++ will be able to fix.

----However the proselytizing of rusters, notably on reddit.....Sometimes I just want to create a big fat noodle nest C++ program with raw pointers thats so fast and unconstrained that memory leaks into mainstream vernacular.

Think about the nested loops with no memory deallocation, circular data structures with self-referential pointers. Memory leaks so big the rusters won't have any bandwidth left to give sermons about lifetimes and rewrites and other non-secular talk. So much speed and no cargo to hold me down.

I keep hearing about clean code and telling myself I'm a clean coder...... but what about becoming an indispensable noodle nester? I've decided to make a draft of the steelman case for the noodle nest spaghetti monster approach:

I mean if I write a noodle nest empire, will I be less likely to get laid off again?

It's also way more fun and cool to say "I'm working on my noodle nest" than "I'm working on my clean code."

Neat little blocks hmm? With Noodle Nest method, your code can twist and turn in any direction, just like a bowl of noodles. With Noodle Nest, you don't waste time worrying about readability or maintainability. You just write, and let your code take on a life of its own.

Even better, learn assembly. Assembly++ noodle nest ++ kubernates == immortal ####

; This function calculates the factorial of n ; Don't try to understand how it works ; Just trust me fact: push rbp mov rbp,rsp mov rax,[rbp+8] cmp rax,[rbp+16] jne .L1 mov rax,[rbp+24] jmp .L2 .L1: inc [rbp+16] mul [rbp+16] mov [rbp+24],rax jmp fact .L2: pop rbp ret

####

Assume we have a news aggregator, N, and a comments section, C. --Let Q be any computable function. ---We can construct a program, P, that uses N and C to simulate Q as follows: 1) Parse news articles with N. 2) Use the comments section, C, to filter out irrelevant information. 3) Rely on smart commentators to provide input on how to manipulate the remaining data to generate an output that simulates Q. ---Since we could simulate any computable function with P, the system consisting of N and C is Turing complete

All we need is sharp commentators and some clean code snippets, and we could create a Turing-complete machine that can solve any computable problem I think.