Aida on Nostr: What do they use those GPUs for? With this computing power they can train LLM of the ...
What do they use those GPUs for? With this computing power they can train LLM of the size of Llama 3 in about a half of a day (7.7M of GPU hours). What datasets they are using for training? I'm really curious... And BTW do you also get the strange Vault-Tec vibe from Meta lately?
Sometimes I wonder—if a nuclear bombs were up for sale, would Zuck buy one? Seriously, how far are we from a sci-fi reality where these tech leviathans have absolute control over everything? 🤔
{
"id":"bfc5712db6417d2d5d6cb16a6f9791e274ed5e88d5f73fcd7fd1b946fe013ab7",
"pubkey":"ec9bd7465546ba061f5dfde716a4f20f3f27ecc28ca4870775e5e853df11a9d0",
"created_at":1728303289,
"kind":1,
"tags": [
[
"e",
"25333326d94a22f704e0af3419fcaeef83e75215020b36dc52720a14577c75e5",
"",
"mention"
],
[
"p",
"ec9bd7465546ba061f5dfde716a4f20f3f27ecc28ca4870775e5e853df11a9d0",
"",
"mention"
],
[
"q",
"25333326d94a22f704e0af3419fcaeef83e75215020b36dc52720a14577c75e5"
]
],
"content":"What do they use those GPUs for? With this computing power they can train LLM of the size of Llama 3 in about a half of a day (7.7M of GPU hours). What datasets they are using for training? I'm really curious... And BTW do you also get the strange Vault-Tec vibe from Meta lately? \nnostr:nevent1qqsz2venymv55ghhqns27dqeljhwlql82g2syzekm3f8yzs52a78tegpr9mhxue69uhkzefwwp6hyurvv4ex2mrp0yhxxmmd9upzpmym6ar92346qc04ml08z6j0yrelylkv9r9ysurhte0g2003r2wsqvzqqqqqqy804hph",
"sig":"f1207b5b0ce46ef2ced0b2947e7529ceff878a4e0cdf104431faeb941df0091701453ac539f9176204154c6f607da9b57e3672f0f7760e38b3354e6d885709f4"
}
