hypermemetic

836 posts

hypermemetic banner
hypermemetic

hypermemetic

@hypermemetic

forever in the unknown

Katılım Eylül 2024
262 Takip Edilen45 Takipçiler
Sabitlenmiş Tweet
hypermemetic
hypermemetic@hypermemetic·
look ahead into the unknown future, passed all easy guesses, and pull out a token
English
0
0
0
140
brew
brew@0x62797465·
@alexocheema that's like comparing a PS1 GPU to a Blackwell 6000
brew tweet media
English
1
0
75
4K
Alex Cheema
Alex Cheema@alexocheema·
My M4 Max MacBook gets 3,756,165 tok/sec in pure C, compared to ~50,000 tok/sec with the FPGA. Try it yourself: github.com/AlexCheema/tal…
luthira@luthiraabeykoon

We implemented @karpathy 's MicroGPT fully on FPGA fabric. No GPU. No PyTorch. No CPU inference loop. Just a transformer burned into hardware, generating 50,000+ tokens/sec. The model is small, but the idea is not: inference does not have to live only in software 👇

English
67
99
1.7K
230.5K
Codetard
Codetard@codetaur·
your "claude number" is the version of claude required to peel you off from reality. yes, it's funny that richard dawkins' claude number is only 4.7, but you have one too. maybe it's 12. but maybe its 4.8.
English
146
184
4.8K
240.1K
hypermemetic
hypermemetic@hypermemetic·
in any sufficiently expressive system, you can get a sentence that talks about itself
English
0
0
1
11
hypermemetic
hypermemetic@hypermemetic·
some maps model the system which contains them
English
0
0
0
9
hypermemetic
hypermemetic@hypermemetic·
the territory contains the map
English
0
0
0
7
hypermemetic
hypermemetic@hypermemetic·
The universe is as the universe does
English
0
0
0
7
hypermemetic
hypermemetic@hypermemetic·
ai'm clauding today
English
0
0
0
5
hypermemetic
hypermemetic@hypermemetic·
@mmkostrzewa @hackerdocc we could get into all kinds of mysticism here. good stuff. I'll be considering this. What the universe is like to you is largely dependent on which features of the universe are enticing to you.
English
0
0
0
16
Marcin Kostrzewa
Marcin Kostrzewa@mmkostrzewa·
@hypermemetic @hackerdocc To be fair: I have not intended to give any positive answers here. My whole point is that "the universe is like..." is a Rorschach test, the answer is all about the mind of the person answering and has very little to do with the universe itself.
English
1
0
1
12
Eduardo
Eduardo@hackerdocc·
i'll bite the bullet: the men who thought the universe was a clock/mechanical machine were not that fundamentally wrong. now we definitely know that the universe is indeed a computer instead, in a sense that will be even less fundamentally wrong once we realize it's not.
English
1
0
6
727
hypermemetic
hypermemetic@hypermemetic·
@mmkostrzewa @hackerdocc The universe is like a clock, a computer, a tapestry woven from countless threads, the weaver, it's like infinity and the uncountable, it's like a mind arising emergent from the relationships between all matter, and many smaller minds collaborating and battling, and more.
English
1
0
0
25
Marcin Kostrzewa
Marcin Kostrzewa@mmkostrzewa·
This is a dead end – I keep saying "there is no reason for the universe to appeal to contemporary human sensibilities" and you keep responding with "right but I think it's like this because that would appeal to my sensibilities". Humans live in a system of loosely connected tribes and clans -> "the universe is governed by a system of clans and tribes of gods, I'd be very surprised if it was otherwise"; Humans move to a system of autocratic rulers -> "the universe is governed by a single, all-powerful god, I'd be very surprised if it was otherwise"; Humans invent machines that can perform certain tasks more precisely and with more strength than humans -> "the universe was created by a god, but it functions without intervention, like a clockwork, I'd be very surprised if it was otherwise"; Humans move to more decentralized governance, computation and systems thinking -> "the universe is decentralized and all phenomena are emergent from simple computational rules, I'd be very surprised if it was otherwise". So I'm now facing a choice: either we have a tendency to project our material conditions onto our belief structure, or we happen to be the generation that got the ideas exactly right and future generations won't look at us like we're a bunch of babbling baboons.
English
2
0
1
31
Siddhant Nitin
Siddhant Nitin@siddhant_8019·
Every startup founder is struggling. @karpathy says: You can outsource your thinking, but you can't outsource your understanding
English
3
0
2
69
hypermemetic
hypermemetic@hypermemetic·
@mmkostrzewa @hackerdocc Thank you for putting so much effort into untangling the dogmatic universalisms. You helped release my mind from a crystallization
English
0
0
1
15
Marcin Kostrzewa
Marcin Kostrzewa@mmkostrzewa·
A major problem is that we're meatsacks shaped by a very thin slice of the universe across pretty much all dimensions that we know of: size, duration, energy, gravity, EM wavelengths. We have also through great effort managed to peek somewhat beyond that, by magnifying what is too small and shifting wavelengths towards what we can see. Even in what little we've managed to peek, we keep finding horrors beyond imagination, literal Lovecraftian madness-inducing shit. QM is a great example - we can work with it symbolically and even have some success in predicting the future this way, but we don't *get* it, we struggle, the thing feels wrong and foreign, because it is so far outside of the meatsack conditions. We also keep falling on metaphors, like the atom being a bunch of tiny balls flying around, because the truth just doesn't fit our brains. The modelling capabilities of computers are also laughably bad if you think about it. They are impressive to us because they still let us achieve cool stuff, but also most computations are paywalled either by chaotic dynamics or NP hard problems. "Drop a glass on the floor, what will be the shapes and locations of the pieces" is a completely intractable problem, which does not instil confidence in our power. So I find your main premise unacceptable: I see no reason why "what fits our experience" and "the best abstraction we could imagine" would a priori be anywhere close to how stuff really operates. Also it is entirely unsurprising, as humans surrounded by machines, operating in a society focused on machines and constantly thinking about machines, that we'd find the idea of a machine a good and comfortable (as in, snuggly fitting our experience) model for other stuff.
English
4
0
0
43
hypermemetic
hypermemetic@hypermemetic·
@mmkostrzewa @hackerdocc yeah I think there is a fundamental difference between the computers we use and the substrate on which we've built them (the universe)
English
0
0
1
18
Marcin Kostrzewa
Marcin Kostrzewa@mmkostrzewa·
@hypermemetic @hackerdocc If doing something would take you more than the age of the universe and this other thing does it 10^50 times every 10^-10 seconds, maybe it's time to admit there may be a fundamental difference?
English
1
0
1
19
hypermemetic
hypermemetic@hypermemetic·
@mmkostrzewa @hackerdocc When I discuss computation I'm careful to mean, and sometimes say, theoretically. Quantum effects get out of hand quickly
English
0
0
0
13
hypermemetic
hypermemetic@hypermemetic·
I'm with you. I wouldn't say the universe is only one thing. I would hold that it's more like all of those things, and that the analogies only serve their purpose. The universe is computable, but it's likely not in a computer. There is definitely an edge to what we can understand through each analogy, even computation.
English
0
0
1
18
hypermemetic
hypermemetic@hypermemetic·
holographic cli compile a client cli from backend schema, with styling applied to the outputs. Project a real cli from the image of behavior.
English
0
0
0
14