Live data from Hacker News

Bard’s latest update: more features, languages and countries

blog.google

131–140 of 205 posts

Re: Bard’s latest update: more features, languages and countries

#131

This release is GREAT for 1 use case: is to do language learning for KIDS! Since it supports other languages , i can have my young ones go at it talking favorite pokemon, etc. But in a foreign language. That is great. Now if it supported mic input on a foreign language (that may be an edge browser microphone limitation) they could have a full conversation, which would be incredible!

I'm curious, do you have any concerns in unleashing LLMs in front of your kids?

I ask this as someone with no kids who is working on a language learning app that does essentially what you're talking about [1]. We have only done a little bit of red-teaming, but I'm sure there are ways kids could talk about topics that at least some parents may not want.

1. https://squidgies.app

Re: Bard’s latest update: more features, languages and countries

#132
post #49

Earlier quoted context omitted.

Everything google offers is in beta and remains that way for a very longtime. Expect bugs but not regression of core product offerings.

pretty sure youtube/adsense/adwords/google/android/chrome aren't in beta everything else at Google is pretty much an internal startup and insignificant on their numbers

Nope, GCP is now bigger than youtube.

Re: Bard’s latest update: more features, languages and countries

#133
post #130

Earlier quoted context omitted.

*WHAT PRIVACY LAWS*? Canada has incredibly weak, close to nonexistent privacy laws. Google didn't even claim what you're claiming, and instead cited "regulatory uncertainty", which is farcical. It's the same play oil companies do when they want their minions to declare that we should tax them less and give them more.

Which means uncertain about using data collected or uncertain about types of data collected. Oil companies are one of the most taxed industries. Google is not an oil company either. These country restrictions came out months before the bill. They are completely unrelated.

>Oil companies are one of the most taxed industries

Norway benefits 10-fold more from their oil reserves than Canada does. Oil companies have been spectacularly profitable exploiting Canada, yet still there are "woe are the oil company" sorts that are easily leveraged.

I hardly want to talk about oil companies, but it's a perfect example of where Canada being a small country that is easily exploited by mega multinationals comes into play.

>These country restrictions came out months before the bill. They are completely unrelated.

Google originally restricted BardGPT to just the US and the UK. They just opened it up to the world, barring only a couple of embargoed countries...and Canada. Do you understand the discussion you're actually taking part in?

Re: Bard’s latest update: more features, languages and countries

#134
I find the image results to be rather impressive. Or to be more precise, I feel like Bard is much much closer to changing "search" than chatGPT or Bing given the results I'm looking at.

To give an example. I gave it an image of a television listing from some time in 1996. It was for a television film fashioned from a two part episode of The Commish called Father Image. It correctly got all the information from the image, identified it as a TV movie based on an episode of a television show, and most impressively I think claimed that the tv show played on December 7, 1995. That wasn't when the listing was from, but that is the air date of the episode of the television show.

It is clearly using other bits of information (IMDb information about air dates I imagine) and effectively doing multiple db fetches / searches in the background to flesh out the information it is giving back. That is precisely what I want from a "new" style of search. Give it something simple and give me a bunch of curated information which otherwise would have taken multiple Google searches / diving into links to determine. Fun stuff.

Re: Bard’s latest update: more features, languages and countries

#135
post #8
post #6

I tried Bard and it feels worse than GPT-4, it doesn't follow the prompt very well. Uploading an invoice it failed to extract data, instead it responded with hallucinations.

Yes, I think it should be compared to GPT 3.5. Google doesn't really have a public contender to GPT 4 yet but this is what the Gemini project is intended to be, by the newly formed/reformed Google DeepMind team.

Bard seems worse than gpt-3.5 in my experience.

Re: Bard’s latest update: more features, languages and countries

#137

Earlier quoted context omitted.

pretty sure youtube/adsense/adwords/google/android/chrome aren't in beta everything else at Google is pretty much an internal startup and insignificant on their numbers

Nope, GCP is now bigger than youtube.

yeah I did forget GCP

Re: Bard’s latest update: more features, languages and countries

#138
post #21

fun story[0] about Bard's June update - implicit code execution (remember? https://news.ycombinator.com/item?id=36229782 ) - the example that they advertised on their blogpost/video demo now no longer works. Here's what I got as of today: https://g.co/bard/share/2580222b6fb2 (screenshot from a couple days ago: https://twitter.com/swyx/status/1678495067663925248?s=20 ) > What are the prime factors of 15683615? > (afte…

You can't really regression test a model this way because the training output isn't stable. Instead you have a huge test suite (you could even think of it more as an eval benchmark) and you try to make a determination of whether in aggregate you think you're doing better or worse. As a more concrete example, imagine you're training an image classifier, and you have a bunch of ground truth human labeled images. If ver…

yes, as mentioned, im well aware of how evals are done :) https://www.latent.space/p/benchmarks-101

Re: Bard’s latest update: more features, languages and countries

#139

Earlier quoted context omitted.

Don't bother trying their other models available in GCP either. Their embedding models and their un-RLHFd generic GPT analogues are miles behind the competition. It's incredible how bad google have dropped the ball on this. We had some google people come in at $CORPO_DAYJOB the other day to sell their cloud offerings. The engineer had the gall to say "people ask me, why did google miss the boat on LLMs? I say people,…

Yeah, I would say that "we built the boat" is simply a lie given none of the eight authors of the paper still work at Google.

I would say it's still correct.

When a company provides an R&D lab, we still attribute the results of that research to the company as well as to the researchers.

After all, if those individuals were not hired that position would still be filled with someone similar, but it's hard to argue that if the lab didn't bring those researchers together, they may have individually gotten the same result.

Honestly, it's not as easy as one may think to build a research lab.

Re: Bard’s latest update: more features, languages and countries

#140
post #59

Earlier quoted context omitted.

Early search engines would publically show what people were searching for. It was amazing how prevalent silly/embarrassing questions were. Things like "can kissing cause baby?"

There are some fun YouTube videos (old ones) making fun of funny Yahoo! Answers questions that appear to be written by 10-12-year-olds.

How is babby formed? https://www.youtube.com/watch?v=Ll-lia-FEIY
Post reply on HN