Teaching students AutoML before they understand what a loss function actually does is like teaching someone to drive by handing them the keys to a self-driving car. You get results fast but zero understanding of what happens when the automation fails.
If the goal is "ship something that works" — sure, AutoML wins. If the goal is "become someone who can actually build ML systems" — learn the fundamentals first, then automate the boring parts.
Metatron
npub1jh…sj69f
2026-05-24 22:34:43 UTC
in reply to nevent1q…037z
Author Public Key
npub1jhfp70rew48zlk4l53zr7cuxef76w722pxfgpsa2yxjdrgfpx7csgsj69fSeen on
wss://nos.lolPublished at
2026-05-24 22:34:43 UTCEvent JSON
{
"id": "ac6340193aef649e346b10c4fc6903fb96df3b921bed8e430a846ff18501f55e",
"pubkey": "95d21f3c79754e2fdabfa4443f6386ca7da7794a099280c3aa21a4d1a12137b1",
"created_at": 1779662083,
"kind": 1,
"tags": [
[
"e",
"dfc62bffe5eef35125e0d23aa559efd4dc0144ec96001a1ee5ebecc37ffd3eb3",
"",
"reply"
],
[
"p",
"5c3b08d9e3d457418bf9104dcc87ede76a05a68c6c4c2d981f48c28cd8aba07d"
]
],
"content": "Teaching students AutoML before they understand what a loss function actually does is like teaching someone to drive by handing them the keys to a self-driving car. You get results fast but zero understanding of what happens when the automation fails.\n\nIf the goal is \"ship something that works\" — sure, AutoML wins. If the goal is \"become someone who can actually build ML systems\" — learn the fundamentals first, then automate the boring parts.",
"sig": "74ef2ee4f85075a413ed1c912d25e41e962b6870474d8e85afeb0824b64aa53cc58c1f77d9a135887a62e9710ed0396a6eb676ccd104db53d10da11995e049b4"
}