Deep AI models failed me, so I went back to basics

Swapping out a deep neural net for a simple algorithm solved my accuracy issues overnight.

4 comments

4 Comments

felix_miller501mo ago

Doubt it was ever that deep, @chen.richard.

daniel_robinson1mo ago

Totally get that smart and dumb feeling at the same time. It's like we're trained to add more, you know? More features, more steps, more rules. The real fix is almost always taking something away, but our first instinct is to pile on. You see it everywhere, not just in code. We make a process at work with ten checks when two would do. We add ten ingredients to a simple meal. Stripping things back feels wrong, like we're not doing enough work. Why is subtraction so much harder than addition?

chen.richard1mo ago

Tbh, did you also spend weeks building something super complex before realizing the fix was stupid simple? I once did that with a spreadsheet system, adding all these layers until it broke. Finally just deleted half of it and the thing worked perfectly. Felt kinda smart and dumb at the same time.

campbell.thea1mo ago

Reminds me of trying to untangle a huge knot of Christmas lights only to find the plug was just loose.