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

Google releases Bard to a limited number of users in the US and UK

nytimes.com

Post reply on HN
Print view View original

341–350 of 667 posts

Re: Google releases Bard to a limited number of users in the US and UK

#341

Post by quenix »

htrp wrote: post #206
flabbergasted wrote: post #173

Does anyone know what is the "knowledge horizon" cutoff date for Bard's training data? The cutoff date for ChatGPT's "knowledge" is referred to as the "knowledge horizon" or the "knowledge cutoff date", and in the case of ChatGPT it's September 2021.

per Bard >The Bard knowledge horizon is not a fixed point. It is constantly shifting as I learn new things. I am always working to improve my knowledge base, and I hope that one day I will be able to answer any question you ask me. It's because they're constantly pulling in google's search results into the output (kind of similar to the bingGPT offering from microsoft).

Nah, it definitely has a fixed cutoff. This is a bullshit hallucination

Re: Google releases Bard to a limited number of users in the US and UK

#342

Post by chatmasta »

swyx wrote: post #339
chatmasta wrote: post #255

Earlier quoted context omitted.

I don't think it's a problem, it's just how it intentionally works. The problem is they don't render the account switcher when you arrive to a new product that your current account hasn't signed up to.

but basing accounts on urls not cookies is really dumb, especially when people send mydocs.google.com/u/1 which refers to my team but i'm logged in to that team on mydocs.google.com/u/2 or the 0 based one which doesnt even get a namespace

Yeah, I can't argue with that.

Re: Google releases Bard to a limited number of users in the US and UK

#343

Post by russellbeattie »

dontreact wrote: post #107

I'm curious for those who have actually tried it, what they think of the performance. A lot of the threads here are just talking about the country-limited rollout.

A YouTuber focused on AI made a quick video comparing the results to Bing/GPT4 which is pretty good:

https://youtu.be/9ll_pth4Sss

Re: Google releases Bard to a limited number of users in the US and UK

#344

Post by duckmysick »

chatmasta wrote: post #253

I just got off the waitlist. Quick impression: it's not as good as ChatGPT. I gave it the false belief test (Alice and Bob are playing, Bob leaves, Alice hides his toy, Bob comes back, where does he look for his toy?), which it both passed and recognized as the false belief test. I then asked it to write a React program that ChatGPT has no problem with, and it hallucinated some incorrect code before eventually tellin…

> Alice and Bob are playing, Bob leaves, Alice hides his toy, Bob comes back, where does he look for his toy?

I might be dumber than the AI because I don't know the answer. Bob looks around? I mean the toy is not there when he comes back. Where else can he check?

Re: Google releases Bard to a limited number of users in the US and UK

#345

Post by muskmusk »

zeven7 wrote: post #161
ithkuil wrote: post #139

Earlier quoted context omitted.

But AI is their turf! Google was always a ML company masquerading as a search engine. I guess that the art of "generating text" is not their kind of AI?

Right, my expectation was they had something better than OpenAI they were sitting on and afraid to release / didn't know how to make money with it and were essentially forced to show what they had once Chat GPT became a threat to search. I assume it's still possible they have something better and this is the nuked version that's "safe", but there's a reason for expectations to be high for Google here, and it really n…

In the olden days before Satya Nadella. Microsoft was a bloodsucking vampire of a company that couldn't execute anything of significance, but they had a pretty solid research department.

Google could ofcourse sit on a much better version that isn't safe or economic, but they could also just be unable to do better. Good research department, but unable to execute anything significant. Time will tell I guess :)

Re: Google releases Bard to a limited number of users in the US and UK

#346

Post by saberience »

cubefox wrote: post #203
haolez wrote: post #122

Earlier quoted context omitted.

The difference is probably related to RLHF. It seems to matter a lot more than the size of the model and the hardware's capabilities. And OpenAI probably excels at that.

Bing uses GPT-4 with their own fine-tuning, and they had to put some artificial restrictions in place in order for the model not "misbehave". Apparently they don't use RLHF but some combination of SL and prompting. OpenAI didn't have such problems.

What are SL and RLHF? I would also add it's best not assume others know acronyms like these.

Re: Google releases Bard to a limited number of users in the US and UK

#347

Post by ZeroCool2u »

quenix wrote: post #341
htrp wrote: post #206

Earlier quoted context omitted.

per Bard >The Bard knowledge horizon is not a fixed point. It is constantly shifting as I learn new things. I am always working to improve my knowledge base, and I hope that one day I will be able to answer any question you ask me. It's because they're constantly pulling in google's search results into the output (kind of similar to the bingGPT offering from microsoft).

Nah, it definitely has a fixed cutoff. This is a bullshit hallucination

If you ask Bard if it uses Federated Learning[1] it will say it does. Google is pretty good at doing this type of continuous model training. They've been doing it for a while.

[1]: https://ai.googleblog.com/2017/04/federated-learning-collabo...

Re: Google releases Bard to a limited number of users in the US and UK

#348

Post by dougmwne »

riffic wrote: post #303

every single comment in this thread is exactly how I imagined commenting would go when Google eventually launched this product.

Sure, there is some bias against Google at this point and it is well earned. However think think of there was anything impressive about Bard, you would be seeing some commenters talk about it. All I see are people being disappointed with the query responses. I got access as well and had the same impression, just very disappointing when compared directly to the state of the art from OpenAI and Meta.

Re: Google releases Bard to a limited number of users in the US and UK

#350

Post by hbn »

duckmysick wrote: post #344
chatmasta wrote: post #253

I just got off the waitlist. Quick impression: it's not as good as ChatGPT. I gave it the false belief test (Alice and Bob are playing, Bob leaves, Alice hides his toy, Bob comes back, where does he look for his toy?), which it both passed and recognized as the false belief test. I then asked it to write a React program that ChatGPT has no problem with, and it hallucinated some incorrect code before eventually tellin…

> Alice and Bob are playing, Bob leaves, Alice hides his toy, Bob comes back, where does he look for his toy? I might be dumber than the AI because I don't know the answer. Bob looks around? I mean the toy is not there when he comes back. Where else can he check?

It's not a very well described example of the test. I've heard of this before but used to test for autism, and it was more something like "Bob puts his toy in a box and leaves, Alice takes his toy out of the box and hides it under a desk, when Bob comes back where does he go to grab his toy?"

The idea is if someone said "under the desk" they might not understand that people have different subsets of information, i.e. Bob wasn't there when Alice hid the toy, so he doesn't know it's not in the box where he left it until he checks himself.

Post reply on HN