@six

In most domains a poor quality product may not be actively harmful but rather just less effective at accomplishing its purpose. With models its different though because a worse model can actually be actively harmful vs not using AI at all But now that I think about it this is not unique to the product category of LLMs, actually any product in a domain considered "critical" has this property. like flying in a poorly built airplane is worse than not flying at all but like a bad doorknob is not worse than having no doorknob. It's just harder to get inside but at least you can still get inside. Though maybe you want a push/pull or sliding door instead of a doorknob to begin with Something to think about

1 reply

0 recast

6 reactions