The experiment we just did at Penn to see if AIs can do homeworks for AI students concludes with everyone reassured that AIs completely fail at getting the logic right for anything complex, even with good feedback for improvements
I tried having it generate code for markov decision processes, reinforcement learning, basic language models, and perceptrons, and the AIs did an absolutely terrible job
You can see the homeworks we tested on here http://artificial-intelligence-class.org/
J Λ M Ξ S /
npub1sv…y6mcs
2024-05-18 01:13:09
Author Public Key
npub1sv4r2lnugqz7f98jlwzqx7zz3ugt2hkzpzl2fpc82c7kpfv2rukqgy6mcsPublished at
2024-05-18 01:13:09Event JSON
{
"id": "265075347c031db76a782322fb1c9ef65fb38f0afaabbe1e303285b750411c7e",
"pubkey": "832a357e7c4005e494f2fb840378428f10b55ec208bea48707563d60a58a1f2c",
"created_at": 1715994789,
"kind": 1,
"tags": [
[
"proxy",
"https://mastodon.social/@jmsdnns/112459434522592035",
"web"
],
[
"proxy",
"https://mastodon.social/users/jmsdnns/statuses/112459434522592035",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://mastodon.social/users/jmsdnns/statuses/112459434522592035",
"pink.momostr"
]
],
"content": "The experiment we just did at Penn to see if AIs can do homeworks for AI students concludes with everyone reassured that AIs completely fail at getting the logic right for anything complex, even with good feedback for improvements\n\nI tried having it generate code for markov decision processes, reinforcement learning, basic language models, and perceptrons, and the AIs did an absolutely terrible job\n\nYou can see the homeworks we tested on here http://artificial-intelligence-class.org/",
"sig": "cf02ff81e9d1f5c8595571c3fff2e95ef81c3c414d238487fc04bf9405492fd69a928f759265ddf6f127a7e0864a4d8a29c9a63702c1d292460e7cf287169b06"
}