Waste of energy and increased response latency from useless introductions
#2
by
cmp-nct
- opened
Look at the main example:
"Als KI kann ich keine persönlichen Beobachtungen teilen, aber ich kann einige allgemeine Informationen zur Fahrradwegesituation in Hamburg liefern."
That's 40 tokens of generation without any content, it's like an advertisement before a youtube video about something the user does not care about.
That costs significant wasted performance resources and delays.
To really use such models they would need further fine tune to remove the spam prefix phrases.