Ame on Nostr: The core issue for this, along with a lot of other things, is scale. When the ...
The core issue for this, along with a lot of other things, is scale. When the algorithms that made good search engines possible came into being circa 2010, Google already had a monopoly, and could use hundreds of millions of users' data for A-B testing and training models of which search results were relevant to which users and which searches. You need a critical mass of training data before a large-scale machine learning model becomes useful, and you need to be the best in the business already to get the kind of usage required to train such a model. My fear is that this situation can't be replicated, and that those five or so years in which Google worked will never happen again.
Published at
2023-06-15 11:26:28Event JSON
{
"id": "c21205f1801bb133bcd26fa1dd9e2bbef597c3e1605526f8b5410541735e677b",
"pubkey": "7271f65788cf3d629727a54de8a74cfaa69f3ee1232c7dfc6bcb8059d28467e1",
"created_at": 1686828388,
"kind": 1,
"tags": [
[
"p",
"3d64fdb61082d26d838703d4e0d91d8778759a98196a75e7bb21079aa47596be",
"wss://relay.mostr.pub"
],
[
"p",
"c96655f60dc28762c0e4011034ca01b053586942f05683c6207bf371f5423be1",
"wss://relay.mostr.pub"
],
[
"p",
"748fa8a28d3a079b7650de9088739a8ce36fdbd6d83b004565b1609787bde390",
"wss://relay.mostr.pub"
],
[
"e",
"1ae50ace752d9eeef62180de479413db3f462aed071495f9d03c78530a2a39de",
"wss://relay.mostr.pub",
"reply"
],
[
"mostr",
"https://poa.st/objects/90b7f9d2-7e7c-4429-adef-16cf5a0eb034"
]
],
"content": "The core issue for this, along with a lot of other things, is scale. When the algorithms that made good search engines possible came into being circa 2010, Google already had a monopoly, and could use hundreds of millions of users' data for A-B testing and training models of which search results were relevant to which users and which searches. You need a critical mass of training data before a large-scale machine learning model becomes useful, and you need to be the best in the business already to get the kind of usage required to train such a model. My fear is that this situation can't be replicated, and that those five or so years in which Google worked will never happen again.",
"sig": "40d1ce558249f3727b2b637fc18b35ec62f7bb0195833c12c7eea5d83ac03b79e67346d8031ca0eda6640dbc9063be3072216b552b599875eb18b9a80646eaef"
}