Andrew Gallagher on Nostr: LLMs are the result of training, and they were trained to generate statistically ...
LLMs are the result of training, and they were trained to generate statistically likely sequences of words, not to be logically correct. *Never forget this*
Published at
2024-04-11 08:43:13Event JSON
{
"id": "22f1c26bf4d4c3df49247ca666f1eec1adc563c5883c63b60942d9a9c6dd35f0",
"pubkey": "761b9a2018447d7df5cfd17f4ade117f90e8f2335d5f8012ade48b5d1d0c1480",
"created_at": 1712824993,
"kind": 1,
"tags": [
[
"proxy",
"https://mastodon.ie/users/andrewg/statuses/112251698765538380",
"activitypub"
]
],
"content": "LLMs are the result of training, and they were trained to generate statistically likely sequences of words, not to be logically correct. *Never forget this*",
"sig": "2534b9f6d191a7b1fbad8f522b0eb2bcd53c3c4329c7fb550700c969fdfbe5c52051466c4cb4bb8d5992a0092ecf4537ab81d201ea6d3f1c39c6c1c16fe3e0c2"
}