John Regehr on Nostr: I've been seeing more posts recently about AIs getting poisoned by training on AI ...
I've been seeing more posts recently about AIs getting poisoned by training on AI bullshit, but this reminds me of the time I assigned students to go find a red-black tree implementation on the web and then test it and fix the bugs they found. most of the available implementations were so wrong that the assignment was infeasible, and this was well before the current wave of AI-generated bullshit.
Published at
2024-05-03 18:05:23Event JSON
{
"id": "8f58748fe510cc92c9937995f3bbfb8f060cb594522a55abe420a7871c6a9568",
"pubkey": "a1a4eb540235341a2db0d8a08cfb74edc4bb06b316c7bb39580c55559ce71ba1",
"created_at": 1714759523,
"kind": 1,
"tags": [
[
"proxy",
"https://mastodon.social/users/regehr/statuses/112378480142697397",
"activitypub"
]
],
"content": "I've been seeing more posts recently about AIs getting poisoned by training on AI bullshit, but this reminds me of the time I assigned students to go find a red-black tree implementation on the web and then test it and fix the bugs they found. most of the available implementations were so wrong that the assignment was infeasible, and this was well before the current wave of AI-generated bullshit.",
"sig": "73b5ed48a079e751bf3a03f9283b45b0b05bec9ea9877669e36fe4dec5922d23bbabfdb5f1351e639aa03bd497d3ed0981504e9b438f7b9674e9ca1f0464e256"
}
