My AI training run hit 87% accuracy and I didn't even touch the model

I was cleaning up some old project folders and found a dataset I made two years ago for a basic image classifier. On a whim, I fed it into one of those new autoML platforms, the kind that just needs a CSV file. I set it to run overnight and forgot about it. Woke up to an email saying the model hit 87% validation accuracy. The thing is, I didn't write a single line of code or pick the architecture. It just... figured it out. Has anyone else had a result that good from a completely hands-off process?

3 comments

Log in to join the discussion

Log In

3 Comments

shanejackson6mo ago

Makes you wonder what else is just sitting in old folders...

1

theawest6mo ago

What if most of it's just boring junk though?

2

andrew8546mo ago

But that's the thing, right? If we call it junk and ignore it, we're making a choice about what gets saved. Who decides what's boring now but might matter later? My grandpa's tax records from 1970 seemed like junk until we needed his birth date for something. What if we're throwing away pieces of the puzzle just because we don't see the full picture yet?

4