Couple of thoughts on A.I. before bedtime.
I've seen the argument "these tools are only going to get better," but I'm not sure. Imagine a graph where the X axis is "volume of training data" and the Y axis is "A.I. capability." Right now, lots of money is being bet on that curve being exponential, but there are papers out that suggest it's much more likely to be an asymptote.
In other words, shit don't scale.
And if shit really don't scale, then we've got trouble brewing.
Legit_Spaghetti /
npub1te…x865f
2024-06-19 04:17:48
Author Public Key
npub1tef9l9qzawwlslemyaaaqlwv5x45vzm62kad4xdnuuph27lvcrxsyx865fPublished at
2024-06-19 04:17:48Event JSON
{
"id": "7e9ce03bf18a4520ff883dc2b380a12f8dc2a0388839c78931ef2cb4c8ed3083",
"pubkey": "5e525f9402eb9df87f3b277bd07dcca1ab460b7a55bada99b3e703757becc0cd",
"created_at": 1718770668,
"kind": 1,
"tags": [
[
"proxy",
"https://mastodo.neoliber.al/users/Legit_Spaghetti/statuses/112641354558467739",
"activitypub"
]
],
"content": "Couple of thoughts on A.I. before bedtime.\n\nI've seen the argument \"these tools are only going to get better,\" but I'm not sure. Imagine a graph where the X axis is \"volume of training data\" and the Y axis is \"A.I. capability.\" Right now, lots of money is being bet on that curve being exponential, but there are papers out that suggest it's much more likely to be an asymptote.\n\nIn other words, shit don't scale.\n\nAnd if shit really don't scale, then we've got trouble brewing.",
"sig": "06ff468c2a9f69533dc2c5cffdf61df093ad8175d02c51c3effda8a8a0b93ff393a6f8684e44d23fbb031282d13aa2460a09a203348d2776b440632af4aee497"
}