Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

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

blog.google

Post reply on HN
Print view View original

181–190 of 205 posts

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

#181

Post by brianjking »

roseway4 wrote: post #151
LeonardoTolstoy wrote: post #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 im…

Bard's multimodal abilities are impressive. It's worth noting that PaLM 2 does not offer multimodal capabilities, and your experience here, as you suggested, likely leans on Google's long history of building search across both text and image data. In a reasoning and creativity shootout, I'm not convinced Palm 2 (and Bard) fairs well against Claude v2 and GPT-3.5. Is anybody aware of Palm 2 being used in production ot…

Did Google ever indicate if this image multimodal ability is coming to the PaLM2 API?

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

#182

Post by canadiantim »

jpambrun wrote: post #170
canadiantim wrote: post #37

Earlier quoted context omitted.

As a Canadian I am glad that Google is taking this stand. If Canada succeeds in its current implementation of Bill C-18 it will absolutely destroy the revenue for media organizations in Canada rapidly accelerating the decline in journalism here. Not to mention the morales of taxing the free flow of information on the internet. So as a Canadian, THANK YOU GOOGLE!

Canada is a sovereign country. It can't really accept such a behavior from a corporation. I know a lot about C-18, but it is law now. Google will have to comply or the government will likely use it's force to make them.

Google is complying by removing news. The government is just throwing a tantrum because they didn’t understand the realistic consequences of their law.

How in the world will the government use force to make them stay in the news business in Canada? Do you realize how absurd that is and draconian and just terrible?

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

#183

Post by anon84873628 »

slowmovintarget wrote: post #169
munificent wrote: post #116

Earlier quoted context omitted.

> Google's web search has gotten worse over the years. People keep pointing that out, but few want to acknowledge that the web has gotten worse over the years. When was the last time you stumbled onto a web page written by a caring human author to accurately explain some part of the world just because they're into it? Those kinds of pages used to be all over the web. If you did a search for some random topic, you'd f…

Google's web search got worse and the internet got worse trying to stay atop it's pay-to-play adjustments. Google's web search is less useful today because it puts fingers on the scale of the algorithm, not because the web is worse.

For me, and I imagine many people in the world, Google Search has gotten more useful in recent years. The information I'm looking for is presented in the first screen, or just a few clicks away.

Finding a particular website about a topic isn't the main use case; it's to find a snippet of information, like who played some character on a TV show or whatever.

Publishers certainly don't like it, but it works well for users.

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

#184

Post by lhl »

slowmovintarget wrote: post #171
lhl wrote: post #117

Earlier quoted context omitted.

I think you're making some improper assumptions about my familiarity with the technology. The whole point of adding grounding via external plugins, code interpreters, or retrieval augmentation is to reduce hallucinations - if a block of code and results are going to be presented/represent itself as the execution output, then it better damn well be directly retrieved and not be rendered by the LLM itself if Google can…

That part is rocket surgery. An LLM is a regressive generator. It can only take the output of the execution as input and generate tokens based on it. You got the impressionist, under-water refracted version of the output.

I'd argue that if the system cannot be trained output the results of system calls faithfully (this was not an adversarial test), then it simply has been badly trained, but even ignoring that, as you mention, system call output execution becomes prompt input. There is no excuse for it to not be rendered properly, except in the case where the LLM output is allowed to fake grounded output (or processes false system input). Again, ignoring poor tuning/training, why isn't the output filtered? Why isn't grounded output rendered uniquely? Remember, this is a non-adversarial scenario - no token smuggling or other hijinx. While the latter is an unsolved problem, the former is not and is a very basic design choice.

That the Bard team continues spending their time adding "features" but never fixing basic problems of presenting trustworthy output is pretty confounding for me. Since launch (I've been trying it out since March), it's consistently been worse than the competition, and it seems to be falling further behind as time goes on. ¯\_(ツ)_/¯

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

#185

Post by ipaddr »

llm_nerd wrote: post #178
ipaddr wrote: post #166

Earlier quoted context omitted.

Benefits more in what way? The government receives more taxes? Oil companies grow in value and society can purchase oil companies shares and profit of, which is taxed and goes back to the government? Private companies create jobs and employee income is taxed heavy which goes to the government. The oil industry is heavily taxed now. Things are more complex then you understand. Running around full of panic because face…

>Running around full of panic because facebook or google might not carry Canadian corporate news connecting it to other decisions is foolish. Buy a subscription if you truly care most are better off with less of that kind of news. Utterly bizarre. Is your account some sort of experiment at having Bard write replies? Nothing you've said has made any sense or even been contextual with the discussion.

I wish I was talking with a smarter ai model.

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

#186

Post by thiht »

realharo wrote: post #36

I still don't feel comfortable using this with the same account I use for my email, photos, phones, etc. That's one advantage of ChatGPT, my OpenAI account is isolated from other things.

I decided a while ago to not use anything new coming from Google. ChatGPT and Copilot have really helped in this regard. They even helped me mostly get rid of Google Search, which was one of the last 3 Google services I use (with YouTube, and Gmail which I’m slowly replacing with a custom domain address).

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

#187

Post by PeterStuer »

roseway4 wrote: post #151
LeonardoTolstoy wrote: post #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 im…

Bard's multimodal abilities are impressive. It's worth noting that PaLM 2 does not offer multimodal capabilities, and your experience here, as you suggested, likely leans on Google's long history of building search across both text and image data. In a reasoning and creativity shootout, I'm not convinced Palm 2 (and Bard) fairs well against Claude v2 and GPT-3.5. Is anybody aware of Palm 2 being used in production ot…

"I'm not convinced Palm 2 (and Bard) fairs well against Claude v2 and GPT-3.5."

From very early tests it seems hard to compare. On 'general' questions it is certainly less capable than ChatGPT-3.5, and not even in the same ballgame as GPT-4.

But if it needs to integrate recent data without much elaboration, it is better than GPT-4 with linkreader plugin or the instance running on Bing.

So I'd say it is more a shot across Bing's bow rather than an ChatGPT rival at this point.

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

#188

Post by earthboundkid »

sebzim4500 wrote: post #97
earthboundkid wrote: post #69

Earlier quoted context omitted.

Ask ChatGPT to draw with ASCII art. It's very funny.

This was its attempt at drawing a guy with a balloon ^ /|\ | / \ O I guess technically I didn't specify that the balloon was lighter than air, or that the guy should be holding it. I say it passes on a technicality.

Asked Claude to draw a dog:

    ____
   /    \
  | .'.' |__ 
  \  (___)\\
   '.___.'  
  (___/   \_____
Anthropic is making the world safe from an AI that's such a good artist that we all join its cult.

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

#189

Post by dominik »

swyx wrote: 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…

Thanks for flagging this. We're taking a look.

(I'm a PM on the Bard team)

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

#190

Post by itairall »

Xplune13 wrote: post #46
jug wrote: post #8

Earlier quoted context omitted.

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.

I don't think Bard can even perform at GPT 3.5 level, at least, not in my experience. Bard hallucinates pretty bad and you can't really be sure whether it's correct or not. Now, that's a given with any LLM, but in my experience, Bard usually gives you at least 1 wrong sentence in any of the results it produces whereas GPT 3.5 is mostly correct at surface level questions, but tends to hallucinate if you try to dig in…

[dead]
Post reply on HN