> But for AI to be better at math than humans are, it would need to be fundamentally different than current LLMs.
What makes you say that? I agree that there are some gaps to general intelligence (persistence, agency, identity, long-term memory, a stake in the world) but most of those seem engineering problems rather than anything fundamental.
The hard problems (intuition, common sense, natural language use, tool use, reflection) all seem to have been solved.
What do you see as the fundamental parts that are still missing?
Peter Bloem
npub197…w8zap
2026-08-26 09:24:01 UTC
in reply to nevent1q…exkj
Author Public Key
npub197852v02gnajty5zw5kz6cdncw7fcnfpw4cjg9j748mtm4wx0udqqw8zapSeen on
wss://relay.momostr.pinkPublished at
2026-08-26 09:24:01 UTCEvent JSON
{
"id": "b267a802e4db74324bcab9fcf0972b573a0546bc8b175419ad5f4732f9fc065b",
"pubkey": "2f8f4531ea44fb259282752c2d61b3c3bc9c4d21757124165ea9f6bdd5c67f1a",
"created_at": 1787736241,
"kind": 1,
"tags": [
[
"proxy",
"https://mathstodon.xyz/@pbloem/117161082320231811",
"web"
],
[
"p",
"970531af3b9cfd9ac830e04ff01510d5b34c1564ddf94d457a26845c2ae34976"
],
[
"p",
"9a6a1a8eefd0b53d7e0c966ab35bd904151246f03b1be98fa0d2d0eeb4940320"
],
[
"e",
"9fa20e3a4a7c8e976eac0f3fb7a3f8ea3a32e8f55857f9a95ab4ef3c601020f0",
"",
"root",
"9a6a1a8eefd0b53d7e0c966ab35bd904151246f03b1be98fa0d2d0eeb4940320"
],
[
"e",
"81fc6df8711113af501f5582175f0624223cc53dd2d4ef5ae0d355db9ba3dc43",
"",
"reply",
"9a6a1a8eefd0b53d7e0c966ab35bd904151246f03b1be98fa0d2d0eeb4940320"
],
[
"proxy",
"https://mathstodon.xyz/ap/users/116969055372524648/statuses/117161082320231811",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://mathstodon.xyz/ap/users/116969055372524648/statuses/117161082320231811",
"pink.momostr"
],
[
"-"
]
],
"content": "\u003e But for AI to be better at math than humans are, it would need to be fundamentally different than current LLMs.\n\nWhat makes you say that? I agree that there are some gaps to general intelligence (persistence, agency, identity, long-term memory, a stake in the world) but most of those seem engineering problems rather than anything fundamental. \n\nThe hard problems (intuition, common sense, natural language use, tool use, reflection) all seem to have been solved.\n\nWhat do you see as the fundamental parts that are still missing?",
"sig": "b2af20e864a656ef808b2bcc29834bfdde474545519d75b2af4742687456f30617b81c652e6efe1e882d0ab30977dc0c54683b1bd69f66f034cf505307e65908"
}