4 ms·
Bard is no where near ChatGPT quality. Not even GPT-3.5. My test was to translate lines of Korean text. The prompt is simple. "Translate these sentences from K
by thisisit 3y ago
Bard is no where near ChatGPT quality. Not even GPT-3.5.
My test was to translate lines of Korean text. The prompt is simple. "Translate these sentences from Korean to English: <Korean text>". ChatGPT works all the time. Quality isn't great but it works.
Bard, on the other hand, gets confused. It simply repeats the Korean text. I need to prompt it again - "I asked for a translation". And then it starts to work correctly.
- 10xDev 3y agoGive us the Korean text. No point concealing it so we can try it for ourselves.
- thisisit 3y agoThere is no hidden korean text. The point was that Bard fails to translate any Korean text in one go.
- numpad0 3y agoThat’s funny given that OpenAI architecture GPTs, even LLaMA-7B, just do it right even language specification is wrong or reversed(they don’t care the exact letter of instruction asks to translate a piece from a language or to a language!)