turmeric_root's comments

On: The Persistent Myth That Most Americans Are Misera...

[turmeric_root]

> Isn't bragging the main thing people do on Social Media?

start following more interesting people

On: Saying Goodbye to OpenSubtitles.org API

[turmeric_root]

whisper doesn't seem to support diarization (identifying when the speaker changes), which is needed for subtitle formatting.

On: Recursively summarizing enables long-term dialogue...

[turmeric_root]

c'mon just write a function that takes in text and tells you whether or not it's true, how hard could it be

On: libjpeg-turbo 3.0 has been released, and why there...

[turmeric_root]

i think it'd be cool if you could make art and still own a house

On: Reddit Threatens to Remove Moderators from Subredd...

[turmeric_root]

> When users join a sub it's because they like the content and tacitly approve of the moderation.

I disagree, I see tons of people blaming admins for moderation actions and vice-versa.

On: Software developer gets $5M judgement against My P...

[turmeric_root]

clearly the arbitration panel ruled against Lindell because it's part of the deep state. /s

On: MiniGPT-4

[turmeric_root]

Windows reserves a certain percentage for VRAM for some reason. So I'd recommend Linux. Or find a way to disable the desktop/UI in Windows.

On: Ask HN: Why don't smartphones encourage programmin...

[turmeric_root]

Have you tried writing code on a smartphone?

On: Show HN: Debate Devil – AI debating practice app

[turmeric_root]

this seems like it might be useful for arguing on HN

On: Sam Altman: OpenAI is not training GPT-5 and "won'...

[turmeric_root]

Worldcoin price skyrockets => cryptocurrency speculation returns => GPU shortage prevents further training of LLMs

[kolinko → turmeric_root]

GPU shortage ended not because of a crash in crypto, but because Ethereum switched to PoS and ditched GPUs. Other cryptos offer just a fraction of payouts from mining, so it's not profitable to mine them on GPUs unless you have free electricity.

[PeterisP → turmeric_root]

Literally none of the LLMs we're talking about were trained on consumer GPUs where the market shortage mattered, they used things like nVidia A100 pods or custom hardware like Google TPU clusters.

The GPUs which are good for cryptocurrencies are decent for using ML models but are not good for training LLMs and vice versa, as the hardware requirements start to diverge. Training LLMs requires not only high compute power but also lots and lots of memory and extremely high-speed interconnect for large models, which ends up costing far more than the pure compute cryptomining needs, making it not cost-efficient for mining.

[adventured → turmeric_root]

There will never be a significant shortage for companies with billions of dollars to spend.