Well, for one thing an #LLM is laughably simple. We think they're complex because of the 100s of Terabytes but compared to the complexity of the real world that's *fart noise*, even before we include subjective decision-making, a process so profoundly involved we literally have no idea how it even happens.
In fact, considering that intelligence is precisely the ability to function on limited & dynamic information, #MachineLearning isn't learning at all.
https://mstdn.ca/@dyckron/113499857086610822
JW Prince of CPH, Radicalized /
npub1qz…s440v
2024-11-17 20:50:47
Author Public Key
npub1qzuu60clkx0ahzwu460av4544xpx3q5k7ckrrt2e4q2dcvz4cnvsqs440vPublished at
2024-11-17 20:50:47Event JSON
{
"id": "ea8619661c40335d399587415383a41dbe868275964b328dfa43c1f01e0b9f77",
"pubkey": "00b9cd3f1fb19fdb89dcae9fd65695a982688296f62c31ad59a814dc3055c4d9",
"created_at": 1731876647,
"kind": 1,
"tags": [
[
"t",
"llm"
],
[
"t",
"MachineLearning"
],
[
"proxy",
"https://helvede.net/users/jwcph/statuses/113500267960081113",
"activitypub"
]
],
"content": "Well, for one thing an #LLM is laughably simple. We think they're complex because of the 100s of Terabytes but compared to the complexity of the real world that's *fart noise*, even before we include subjective decision-making, a process so profoundly involved we literally have no idea how it even happens. \n\nIn fact, considering that intelligence is precisely the ability to function on limited \u0026 dynamic information, #MachineLearning isn't learning at all. \n\nhttps://mstdn.ca/@dyckron/113499857086610822",
"sig": "461a149f9e4159168ff16779b0cd8cdf1c87d726f1c0ae60d2a37c795277c0c12e54124f262a5ffa9e007c1b7d95aecddedaa8e03113d2e0af67aaa2ca630dc2"
}