from Hacker News

Ask HN: How good is Google's Bard?

by FailMore on 2/15/23, 12:57 PM with 3 comments

I'm wondering if anyone has used it and how it stands up against GPT-3/4? For example, with GPT it's possible to have a "conversation" where the previous inputs/answers inform the current one. Is that possible with Bard?

I asked this question yesterday but it was removed from the Ask HN page despite a number of upvotes [1]. If I am misunderstanding the rules of HN/there is a good reason for this to be pulled, please get in touch with me mods. Email is on profile. Thank you.

[1] https://news.ycombinator.com/item?id=34773228

  • by mtmail on 2/15/23, 1:21 PM

    > please get in touch with me mods

    Moderator's email address is hn@ycombinator.com They don't read every post. The old URL is not deleted just the upvote/answer ratio is probably too low for the Ask HN page. It looks like nobody has an answer and Google employees rarely speak of internal tools publicly. I wouldn't read too much into that.

  • by sharemywin on 2/15/23, 1:35 PM

    Large transformer models can look back about 4000 tokens which is about 3000 words.

  • by sourcecodeplz on 2/15/23, 1:00 PM

    Any user that would reply with actual knowledge would be a signal that he/she is/was working at Google.

    I think that is why it got deleted, how can we comment on something we don't have access to?