3 ms·

The responses quoted seem incredible, and are a huge leap in comparison to other chat bots. Perhaps too good to be true? Has anyone tried this? Is there a way

• popcube 5y ago
  author explained that: >LaMDA dialogs reproduced here have any hyperlinks silently edited out. As elsewhere, other elisions for conciseness are indicated by […]. While anecdotal, these exchanges aren’t in any way atypical. However, the reader shouldn’t come away with the impression that all exchanges are brilliant, either. Responses are sometimes off-target, nonsensical, or nonsequiturs. Misspelled words and incorrect grammar aren’t uncommon. Keep in mind that, unlike today’s “digital assistants”, large language model responses aren’t scripted or based on following rules written by armies of programmers and linguists. but even they said this, display record just for make people feel surprised and trust author...