Maybe this simplification will help some people: Modern LLMs (Large Language Models) *interpolate* based on training data that is locked in time. They can uncover sometimes surprising connections that weren't "seen" before, but did exist already. What they can not do is *extrapolate*, as in creating new knowledge, what we humans call creativity, inspiration, lightbulb moments. Thus LLMs are like The Matrix, an artificial virtual world that cannot make progress. It's stuck in time
#SundayThought
Jan Wildeboer 😷:krulorange:
npub1m7…d2e7k
2026-06-28 11:22:47 UTC
Author Public Key
npub1m7m3f90p4y6ff94jyh5swxx9de2m3anangf79md92n96a5m4dx5qqd2e7kPublished at
2026-06-28 11:22:47 UTCEvent JSON
{
"id": "7a62117388993732d0933555c114779496acfa70eef5e22faea851d641e35af9",
"pubkey": "dfb71495e1a9349496b225e90718c56e55b8f67d9a13e2eda554cbaed37569a8",
"created_at": 1782645767,
"kind": 1,
"tags": [
[
"proxy",
"https://social.wildeboer.net/@jwildeboer/116827473050225358",
"web"
],
[
"t",
"sundaythought"
],
[
"proxy",
"https://social.wildeboer.net/users/jwildeboer/statuses/116827473050225358",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://social.wildeboer.net/users/jwildeboer/statuses/116827473050225358",
"pink.momostr"
],
[
"-"
]
],
"content": "Maybe this simplification will help some people: Modern LLMs (Large Language Models) *interpolate* based on training data that is locked in time. They can uncover sometimes surprising connections that weren't \"seen\" before, but did exist already. What they can not do is *extrapolate*, as in creating new knowledge, what we humans call creativity, inspiration, lightbulb moments. Thus LLMs are like The Matrix, an artificial virtual world that cannot make progress. It's stuck in time\n\n#SundayThought",
"sig": "456b73f1fe4300d99d40505f54f7b6242a93ec626619422118ad02ee6c3ad92a3524ad943393b30cbfa9b4da288e6a8937e158fa8e6ccc86c2b443310aefc568"
}