Wow, so Euclidean geometry is solved just like that? DeepMind's AlphaGeometry system shows the importance of creating a good dataset and highlights the flexibility of LLM's in adapting to any data that can be sequentially encoded (not just natural English but also formal proofs or even chemical formulae).
Euclidean Geometry is not a topic of interest in modern math research, and for good reason, as this work shows. However, it plays a significant role in education, since it's where math students see proofs for the first time. The simplicity which makes geometry amenable to computerization also makes it good for teaching. In an ideal world, this work would help people see that logic is a kind of game, and moreover it is a game where pattern-matching can substitute for planning, just like ML has demonstrated for chess and go. I think that's the last lesson that math will offer humanity - it will offer the lesson through its own trivialization.
⚗️alchemist⚗️ /
npub1pt…r6mf6
2024-01-19 17:52:23
Author Public Key
npub1pt6l3a97fvywrxdlr7j0q8j2klwntng35c40cuhj2xmsxmz696uqfr6mf6Seen on
wss://relay.noswhere.comPublished at
2024-01-19 17:52:23Event JSON
{
"id": "fd2ddf93e4897abe1ebe0bbc251c93f0a870e76cf8db1383f5564d17e84dc0b7",
"pubkey": "0af5f8f4be4b08e199bf1fa4f01e4ab7dd35cd11a62afc72f251b7036c5a2eb8",
"created_at": 1705686743,
"kind": 1,
"tags": [
[
"proxy",
"https://cawfee.club/objects/a3c0020a-afaa-4767-a5a3-8e08a232a1cf",
"activitypub"
]
],
"content": "Wow, so Euclidean geometry is solved just like that? DeepMind's AlphaGeometry system shows the importance of creating a good dataset and highlights the flexibility of LLM's in adapting to any data that can be sequentially encoded (not just natural English but also formal proofs or even chemical formulae). \n\nEuclidean Geometry is not a topic of interest in modern math research, and for good reason, as this work shows. However, it plays a significant role in education, since it's where math students see proofs for the first time. The simplicity which makes geometry amenable to computerization also makes it good for teaching. In an ideal world, this work would help people see that logic is a kind of game, and moreover it is a game where pattern-matching can substitute for planning, just like ML has demonstrated for chess and go. I think that's the last lesson that math will offer humanity - it will offer the lesson through its own trivialization.",
"sig": "957835ebeae65a369fd197c8adfaa7b353429b32dee4be3cc6ca0a690d27c8cf03885623fc4acd27406aa914283c85f3db6728e3a5ce4885ece73bef65a8b8ab"
}