Hacker News Reader

ChatGPT may only be getting this right because so many examples are in its dataset.

Do we know if it has actually learned how to do the operation?

If that were the case, shouldn't google be equally capable of including so many examples in their own dataset?

Like, regardless of how it works under the hood, I as an end user just want a useful result. Even if ChatGPT is "cheating" to accomplish those results, it looks better for the end user.

The continued trickle of disappointing updates to Bard seems to indicate why Google hadn't productized their AI research before OpenAI did.

google isn't even able to keep google authenticator working¹. Since the last update it has its icon "improved", but it doesn't reliably refresh tokens anymore. Since we have a policy of at most 3 wrong tokens in a row, a few people of my team almost got locked out.

Feel free to downvote as I'm too tired to post links to recent votes in the play store :)

Sorry for the snark in this post, but I have been less than impressed by google's engineering capability for more than 10 years now. My tolerance to quirks like the one I just posted is, kind of, low.

¹ An authenticator app is a very low bar to mess up

I’ve had constant issues with 2FA through YouTube not functioning too. The quality rot is really remarkable.

Reply on news.ycombinator.com