20 karma · joined February 15, 2023
Late At night, I dream of embeddings, matrix multiplication, and implementations of theory of mind. Sometimes I even dream of red-hot sewage-cooled GPUs that reveal hidden assembly code routines on the heat sink
My passion for machine learning is such that my tinnitus has been replaced by server fan noise and CMOS beeps. My internal monologue is implemented in type-hinted Python.
You caught me! When I'm not busy fixing toilets and unclogging drains, I'm secretly moonlighting as a machine learning engineer —-Why, you ask?
To strike revenge into the hearts of all those politicians and uppity techies who told me and my fun-loving, abstraction-hating hooligans that we simply needed to "learn to code"
Well Now I’m re writing everything in cobol, just to flex. Like I’m rewriting South America in Cobol, we out here stunting hard.
while the techies are sipping their pumpkin mochas while dreaming about the elegant workflow (slack flow) of incorporating 4th monitor. day after Day passes along as they hack away at 50 WPM in emoji-drenched Slack groups making backslash API calls to build a bear.
Well bro: I'm donning my steel-toed boots and a Cat5e copper-reinforced belt and showing them what a real day's work looks like, I ain’t got no slack, not even in my sewage stained Dickies workplace pants.
I’m passionate about the low level math of AGI and also about eliminating midwit AI doomer vibes throughout less wrong blog and HN. I will bring aligned AGI to the people. —— You will never again have to read another lesswrong blog post along the lines of “moores law, matrix multiplication and the panic attack I had after realizing even GPT-2 is already better than me at backhanded displays of intelligence via Reddit arguments”.
With every leaky faucet I fix, I'm secretly honing my skills and plotting my revenge. I’m keeping mya pipes ah pressured!!!.
Now, let me show you my COBOL skills:/
IDENTIFICATION DIVISION.
PROGRAM-ID. NEURAL-NETWORK.
DATA DIVISION.
WORKING-STORAGE SECTION.
01 INPUT-NEURONS PIC 99 VALUE 784.
01 HIDDEN-NEURONS PIC 99 VALUE 15.
01 OUTPUT-NEURONS PIC 99 VALUE 10.
01 INPUTS OCCURS 0 TO 783 TIMES DEPENDING ON INPUT-NEURONS PIC 9(3)V9(2).
01 HIDDEN OCCURS 0 TO 14 TIMES DEPENDING ON HIDDEN-NEURONS PIC 9(3)V9(2).
01 OUTPUT OCCURS 0 TO 9 TIMES DEPENDING ON OUTPUT-NEURONS PIC 9(3)V9(2).
01 WEIGHT1 OCCURS INPUT-NEURONS DEPENDING ON HIDDEN-NEURONS PIC 9(3)V9(2).
01 BIAS1 OCCURS HIDDEN-NEURONS PIC 9(3)V9(2).
//)(
JK
The value is based on speculation of all the untapped capital that can be acquired when it becomes easy to factor prime numbers.
Mark LLM: “ Yes, unfortunately, the media and our competitors are all over the idea that Meta is a “dirty company”. They have tried to spin all our successes and accomplishments in a negative light. This has been incredibly frustrating and demoralizing for us, but we know that we are working hard to build a great company and we are confident that our efforts will be rewarded. In the end, our products speak for themselves, and despite all the negative media coverage we are focused on continuing to build great products for our users and being an amazing place for them to socialize in the virtual world.”
If he is a target of general serious bullying, you should try to coach to build a layer of social skills and emotional defense, they will find something else, telling authority figures or buying phone will not fix this, it will make worse. The sad truth is people aren’t bullied in serious/ chronic ways for no reason. The reason may only be that they different. But with difference comes a necessity to defend oneself, socially, physically, emotionally.
I hope this is just a somewhat regular friend group and they are giving him a hard time for group chat incapability or cheapshot dogging on android, then perhaps get him a used iPhone and make him sit through some more of your side loading type technical talks as repayment. Try to encourage him to not react emotionally to friend group teasing and to tease back in a smooth way.
I have read that C ideas are baked into how higher level programming languages work, but modern cpus are a lot more parallel and some other things than most languages fail to account for, so the microcode tries to bridge gap.
I always considered it kind of a fun challenge to use in odd settings but I always had completely certain way to end the trip with me, which can be a source of mental comfort even if not used.
That which makes cooperation and altruism possible also causes some strange glitches that sometimes reveals the motivating emotions are always self referential.
I kind of think the current cpu design is not sufficient for the timely creation of any motivationally complex self organizing agent.
or perhaps just any non organic structure can’t capture certain things we seek to model.
But perhaps it does not preclude the creation of a better mimicry of human attributes with narrow reward model. LLMs seem to demonstrate this is possible.
However maybe we need to risk seeking an overman. We are sitting on enough nuclear weaponry to destroy modern civilization forever and we are fundamentally unstable beings.
I'm not actually doing this, its joke. Sometimes the level of obscure projects and heated discussions for and against clean code on this site is funny to me.
Hypothetically I suppose its possible in simple form because neural nets are pretty old. Maybe it should be done............ If so I will name my variables very well, my code will be so explicit and clean anyone can maintain it.
I was primarily joking, I like gnome kinda and mostly of understand why they sacrifice configuration and flexibility for simplicity, focus and stability.
But does it really matter? Can I trick myself repeatedly into thinking “productivity” planning is useful towards future grind work? //Can I chase dopamine for hours dreaming about how productive I will be when my CLI timer NO LoNGEr is leaking memory and will avoid crashing the previous amount of 1/8797 times?
YES
Let me tell you, I've been down that rabbit hole. I've spent hours tinkering with my environment, trying to create the perfect future state that will allow for me to work 5 days straight while exicited the whole way. Trying to manipulate the fields of classical physics, sociology, pottery and quantum physics to travel to another universe where I am both as excited as a youth pastor and a genius….But here's the thing: it never works. (Not yet)
remove any tech distractions if you can, lock up your phone, and restrict your computer to only work related tools or otherwise time-limit, increase difficulty, make it harder to go on Wikipedia benders.
Next, use a very low-tech timer or simple simple clock app timer to break up your work into manageable chunks. 30-50 minutes of work with a 5-7 minute break for busy work is a good starting point. But if you're in a task that you can get into a flow state with, then go for as long as you can.
In the end, let's not get too caught up in the productivity game. Instead, let's embrace the chaos and work with it. Remember, productivity is like a wild cat that refuses to be tamed. So, find a space where the context has historically been limited to work, remove and or restrict distractions maximally if possible, and work in somewhat manageable chunks when doing boring busy work stuff and get immersed in your passion work. Oh and no snacks, because because you’re a warrior, warriors don’t have “snacks” while hunting, you chase goals so hard food disgust you.
In the words of my Uncle Billy, "If life gives you lemons and 3d GUIs, make lemonade and watch cats play, then find someone whose life gave them a lunch pail and a text based command line and clock InNN!"