this post was submitted on 23 Apr 2024
549 points (97.2% liked)
Technology
55731 readers
3331 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
As an programmer, I want to think out loud about possible technical solutions.
I would have kept the understandable / hand-made algorithm as the core of search results. If you want to do fancy machine learning, do it on the periphery and we can include the machine output in our algorithm and weight its importance by hand. This would allow us to back out of the decision, because we could lower the weight of the machine learning output as needed.
It sounds like Google jumped strait to including the machine learning in the core algorithm though, and now with a decade of complexity in the core algorithm they are no longer able to go back without huge effort.
In general, it's important to consider "is this a decision we can easily back out of?".
Amazon (and I'm sure others) refers to this as a two way door. Good rollouts minimize impact and can be undone easily.
Exactly, and that's something my company is aggressively moving toward, even though our userbase is nothing like Google's. It's just good engineering to be able to rapidly undo an unfavorable rollout.
Google's operations are absolutely built around the idea of easy rollback. Their products, and the their entire product ecosystem, are not.
Yeah, they seem to do "easy roll-foward." Any service is subject to replacement, given a sufficiently motivated project manager. So if there's a problem in deployment, they just replace the whole thing.