Peter Krupa on Nostr: this was Llama 3.1-70b, but all the other LLMs do the same thing. it's because they ...
this was Llama 3.1-70b, but all the other LLMs do the same thing. it's because they are not actually thinking. they all have been trained on a word puzzle involving a man, a goat, a wolf, a cabbage, and a river, so when you ask a question kinda similar to that, they start reciting something similar to the solution to that puzzle, which is nonsense.
Published at
2024-08-01 17:24:52Event JSON
{
"id": "5b93f232ab8a2888c9d73f0db1d330893ae8a54f53eec5411321fd916e38ab9d",
"pubkey": "358b313c445eccc8485ec27702158b45fa6b768ba7a1bcb99c829ebdc979f27f",
"created_at": 1722533092,
"kind": 1,
"tags": [
[
"e",
"8609d7e4e7ed07de03ecef29c265bbf9545b6fd01c94c5a9a9fe57f6bfe6fe1c",
"wss://relay.mostr.pub",
"reply"
],
[
"proxy",
"https://thepit.social/users/peter/statuses/112887928718122634",
"activitypub"
]
],
"content": "this was Llama 3.1-70b, but all the other LLMs do the same thing. it's because they are not actually thinking. they all have been trained on a word puzzle involving a man, a goat, a wolf, a cabbage, and a river, so when you ask a question kinda similar to that, they start reciting something similar to the solution to that puzzle, which is nonsense.",
"sig": "58ec99d2f472159a029805f43927cbcfc7a14be84f10f0a5a64537f93d5b61b1f88a0c92adf037ed69dd0d30d429f0035a7b90fa275a2b677c31dc6de300f906"
}