I wouldn't say AI is a scam.
1. Not everything that's called AI is a LLM and a lot of these other AI systems (you could also call them statistical evaluation systems) work quite well in their specific use cases (like engine optimization in probably every modern car).
2. Yes, LLMs have some pretty bad properties, like hallucinating things. But on the other side is the first time that there's a computational model that you can actually "talk" with about whatever you like and it will answer in full sentences and with plausible content (even though not correct everytime).
Would I use them in critical use cases right now? No. But nevertheless they somehow work and we have probably not seen the worlds best or most efficient versions of LLMs yet.
Kc
npub1fd…3crha
2024-05-07 20:34:58
in reply to nevent1q…ew2y
Author Public Key
npub1fdf25wzezt27ex9hwvs342yykxnx2m50u6lh59g6dfhk4h69cstsv3crhaPublished at
2024-05-07 20:34:58Event JSON
{
"id": "aae967068614599633d0aa76d6620323e2308488dd1f7a89065dba9d79994dfb",
"pubkey": "4b52aa385912d5ec98b773211aa884b1a6656e8fe6bf7a151a6a6f6adf45c417",
"created_at": 1715114098,
"kind": 1,
"tags": [
[
"e",
"012144e4622488302d378b15f36d444161187272dbe2118e3b3dba47ac04158c",
"",
"root"
],
[
"p",
"9ef9fb2f79d2f506bb0e9943d0534b8fdf95b661d80c889597226ea9b0f3ea6f"
]
],
"content": "I wouldn't say AI is a scam.\n\n1. Not everything that's called AI is a LLM and a lot of these other AI systems (you could also call them statistical evaluation systems) work quite well in their specific use cases (like engine optimization in probably every modern car). \n\n2. Yes, LLMs have some pretty bad properties, like hallucinating things. But on the other side is the first time that there's a computational model that you can actually \"talk\" with about whatever you like and it will answer in full sentences and with plausible content (even though not correct everytime). \n\nWould I use them in critical use cases right now? No. But nevertheless they somehow work and we have probably not seen the worlds best or most efficient versions of LLMs yet. ",
"sig": "35e40b488a05fb11aeb5c0b9e8a5bc3535e73d6399fadbf6af2e0cc32709c3eb387c68bf85d3941ccf25b231dcbe941d72cddc3bb3a5c3c3f25a3787c7fa5d4d"
}