Bard VS ChatGPT

Arf, Bard is not yet supported in France. Hope it will com soon to try it. Seems it should be good but...

max5ty you might be ahead of the game. I just went to sign up for it and was put on a waiting list. (U.S. Resident) The home page says it is only open to US and UK as of now.
For the poster in France, he is probably going to have to wait till Google rolls it out to more countries.



"âœ" You've been added to the waitlist!

Thanks for your interest in Bard. You'll receive an email when it's your turn."

For anyone that's want to sign up - https://bard.google.com/

Yes Chat GPT quite limited in terms of data after 2021. This is what they say themselves:

Limited knowledge of world and events after 2021

I'm yet to try Bard!

Waiting for Bard to release in my region, I have used ChatGPT for now it seems great.

What I Find Interesting

QUOTE:

1) Sensibleness: A measurement of whether an answer makes sense or not;
2) Specificity: Measures if the answer is the opposite of generic/vague or contextually specific; and
3) Interestingness: This metric measures if LaMDA's answers are insightful or inspire curiosity ...

A. Safety: The model achieves a level of safety by tuning it with data that was annotated by crowd workers.
B. Groundedness: LaMDA grounds itself factually with external knowledge sources (through information retrieval, which is search).

/QUOTE
source: https://www.searchenginejournal.com/google-bard/482860/

My Take

From how I understand what that SEJ page is saying -- Sounds to me Google fine-tuned the LaMDA large language model, which powers Bard [ here's the LaMDa paper for more context: https://arxiv.org/abs/2201.08239 ] ...

• And, I think they did this by developing a fine-tuning dataset with columns for input (query), additional_input (harvested formatted content from top organic Google search results for the input), inferencing output from LaMDa model that will be fine-tuned, and scores of inferencing output for sensibleness, specificity and interestingness, in which scores are human-reviewed and annotated ...

My notion is based on similar fine-tuning methods being a common supervised learning technique for large language models (LLMs), which are mostly trained initially through unsupervised learning methods -- OpenAI also used this to fine-tune their GPT models for ChatGPT deployment ...

• Then, I think they added a programmatic step during deployment runtime of this "quantized" (read: reduced model layers, model size and size of weights for optimized performance) LaMDa LLM as Bard's backend, which probably runs something like this:

• User enters query;
• Bard system searches for query in Google, which is set to factor user's personalized and localized searchd data;
• It then harvests content of top organic search results;
• Bard system then feeds user's query and harvested content to model at backend, in order to get model's index of inference outputs; and
• It then gets backend model's inferencing outputs, organizes outputs based on scores for 3 parameters mentioned above, and displays output with highest scores to user.

Implementing Similar Programmatic Steps with OpenAI's GPT

Note: But we can't use GPT4 itself to build a dataset for fine-tuning our own, open source or private LLMs to then use it for a commercial product that competes against OpenAI's products, as this goes against their ToS ...

SERPAPI is used to enter queries in Google and get URLs of organic results. BeautifulSoup4 in Python is used to harvest entire main content section of each page. Llama-Index in Python can be used to build a vectorized library of data during runtime so as to work around GPT4's token limits. So this allows us to feed GPT4 with up-to-date data from Google for it to output up-to-date inferencing outputs for a query (system and user prompts).

• However, what's missing here, and I think this is a really vital factor that can significantly affect LLM output quality, is the user's personalized and localized search data; and
• Using something like SERPAPI to enter a query in Google also means Google won't be able to factor in the user's own search history when deciding which content pages to serve, even if you set Google to user's specific location and device during runtime ...

Anyway, more stuff about accessing Bard dev resources is found below ...

Sample Python Code to Access Bard

$ wget https://raw.githubusercontent.com/La...hatbot/Bard.py
# Source Git repo is https://github.com/LarryDpk/pkslow-samples
$ python Bard.py --session 'UgiXYPjpaIYuE9K_3BSxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx xxx'

Guide to Access Google Bard's API Documentation

QUOTE:

To access the Google Bard API documentation and developer resources, you can follow these steps:

1) Go to the Google Cloud website at cloud.google.com.
2) Click on the "Products" dropdown menu in the top navigation bar, and select "APIs & Services".
3) Click on the "Library" option in the left sidebar.
4) Search for "Google Bard API" in the search bar or scroll down until you find it.
5) Click on the "Google Bard API" card to view more information about the API.
6) Click the "Enable" button to enable the API for your Google Cloud project.
7) Once the API is enabled, you can click on the "Documentation" button to access the documentation and developer resources.

Notes: In the documentation, you can find information about the API's features, how to authenticate your requests, and sample code snippets to help you get started. You can also find additional resources such as guides and tutorials, reference documentation, and community support.

source: https://www.quora.com/How-do-I-get-a...oper-resources

/QUOTE

Github Repo to Access Bard via Python Library

Programmatically Access Bard in Python: https://github.com/acheong08/Bard/bl...CUMENTATION.md

I hope this helps!

Marx Vergel Melencio -

All very interesting...and some of it is over my head.

But, something you said I had thought about and wondered if this was happening:

"Bard system searches for query in Google, which is set to factor user's personalized and localized searched data"

It would seem important that this happens because it seems to me that localized search is what normally happens when someone searches.

If it stops localizing searches it would change a lot of things when it comes to small businesses.

Not sure I said that right, but hopefully you get the idea.

I agree with this. Sometimes chatgpt gives you a little bit wrong information

Although I do think the writing part of the AI platforms is impressive...

some of the other things that AI can do now are even more impressive.

Just for an example, I asked Bard to write me the code I would need to add a countdown timer to a website...I kept this example as simple as I could, but some of the output asking for builds, etc I've read about are mind-blowing.

Read the Twitter post about the guy that built a complete chrome extension in 45 minutes.

(what I asked)

I'm a complete beginner and want to add a countdown timer to my website that counts down to April 30, 2023. Write me all the code I would need to add to my website and explain each step in detail. When finished show me what the output would look like.

(output)

Sure, here is a step-by-step guide on how to add a countdown timer to your website that counts down to April 30, 2023:

Create a new HTML file in your text editor.

Add the following code to the top of the file:

<!DOCTYPE html>
<html>
<head>
<title>Countdown Timer</title>
</head>
<body>
<h1>Countdown Timer</h1>
<p>This timer will count down to April 30, 2023.</p>
</body>
</html>

Add the following code to the bottom of the file:

<script>
var days = 10;
var hours = 2;
var minutes = 30;
var seconds = 0;

setInterval(function() {
seconds -= 1;
if (seconds < 0) {
days -= 1;
if (days < 0) {
hours -= 1;
if (hours < 0) {
minutes -= 1;
if (minutes < 0) {
return;
}
}
}
}

document.getElementById("seconds").innerHTML = seconds;
}, 1000);
</script>

This code will create a countdown timer that starts with 10 days, 2 hours, 30 minutes, and 0 seconds. The timer will then count down to 0 seconds, at which point it will reset to 10 days, 2 hours, and 30 minutes. The timer will update every second, and the current time will be displayed in the "seconds" text element.

Save the file as "countdown.html" and upload it to your web server.
Open the file in a web browser and you should see the countdown timer.
The output of the code will look like this:

Countdown Timer

10 days, 2 hours, 30 minutes, 0 seconds

(end of output)

Of course, I could go back and add specific design features to make the timer more appealing...but you get the point.

The builder I use already has countdown timers, but this is just a simple example of using Bard. If you upgrade to the $20 monthly plan for GPT you can use GPT4 which I understand is better suited than the free version for writing code.

I've been testing Bard on some more advanced plugins, and so far I'm getting good results.

If you're an experienced coder, I'm guessing the questions you would know to ask would probably get way better results than what I'm getting right now.

too bad... Bard is not available in my country
But I really hope to use it in the future

Brandin'wise, sure shit 30+ yeahs gowin' (plus imagry steada acronyms) beats down NEXTO NOOZ MONSTAH fast (ie whatevah).

Question is: that mattah to anywan?

Alternative Question is: this mattah to anywan?

Plus also: where I put muh goddamn hayerdryah?

In the news though:

Eliezer Yudkowsky said at the rate we're going with AI, everyone will die. He's had over 2 decades in AI security.

He was referring to a letter yesterday that elon musk and other top AI scientists had signed that asked for a 6 month pause on AI developments to give the world time to put safety measures in place. He said 6 months was not long enough. They were referring to AI stronger that CGPT4.

He added:

"Many researchers steeped in these issues, including myself, expect that the most likely result of building a superhumanly smart AI, under anything remotely like the current circumstances, is that literally everyone on Earth will die," he asserts. "Not as in 'maybe possibly some remote chance,' but as in 'that is the obvious thing that would happen.'"

Instead, Yudkowsky proposes international cooperation, even between rivals like the U.S. and China, to shut down development of powerful AI systems. He says this is more important than "preventing a full nuclear exchange," and that countries should even consider using nuclear weapons "if that's what it takes to reduce the risk of large AI training runs."

"Shut it all down," Yudkowsky writes. "Shut down all the large GPU clusters (the large computer farms where the most powerful AIs are refined). Shut down all the large training runs. Put a ceiling on how much computing power anyone is allowed to use in training an AI system, and move it downward over the coming years to compensate for more efficient training algorithms. No exceptions for governments and militaries."

So there you go...

tried to do a thread on the 6 month moratorium yesterday but I don't think the thread went through.

For help with creating plotlines, overcoming writer's block, etc. ChatGPT is (so far) better than Bard. ChatGPT plot help is creative; Bard (for naming itself after Shakespeare!) is pretty formulaic.

Just a quick note, I was approved today for Bard. My last post above was when I signed up on the 26. So currently 5 day waiting period here in the US currently.

I am woefully behind on some of this - but from my own reading i think BARD is being developed perhaps in a more thoughtful way than ChatGPT.

I've wondered if by this time next year, there will be a clear choice of 'best to use'...the technology seems to be moving that quickly.

I see some potential 'black holes' but that may be my own level of skepticism. In one test....an AI program recommended 5 books for 'further info' on a question....turns out the books don't exist - they were CREATED by the AI program to answer that question. How could that be of any true value?

My skepticism is based on my own musings of whether AI programs will enable us to be smarter - or allow us to get by with getting dumber by the day....


Two huge technical/scientific fields moving ahead rapidly right now - one in space with the Webb telescope - the other in cyberspace with AI.

bard is still in testing phase i guess

chatgpt started it, and google followed suit.

they're practically the same thing. they're generative AIs trained with tons of data. at first, bard had an advantage, in which it could connect to the internet, whereas chatgpt couldn't.

at that time, on paper, bard should have an advantage, but google somehow failed because bard surfaced wrong data during initial demo.

at this time, i think chatgpt is ahead of bard, and that it's churning lots of data from users who are using it for both work and fun. and also from the fact that microsoft bing is also using it to power its bing chatbot.

but in the long run, i think, bard will catch up.

google has lots of AI projects, and from what it seems, google can put together the many AI projects to power bard.

for example, chatgpt is powered by gpt-4, and this gets updated pretty quickly. but bard which initially uses only lamda, has added palm.

i think, instead of comparing the two, you should try to see which of the two is less prone to hallucination. bing is the worse.

I wonder what happens if you ask Bard to create a search engine algorithm that could put Google out of business?

I'm fascinated to know how iterative we can get with all this.

Gotta tellya, cool to have DanyWorld an' GordonJ an' max5ty chimin' along here.

We all hangin' togethah, we all propulsin' henceforth fromout.

... an' so it goes with intake an outtake.

Or outtake an' intake, yanno.

Either/or, I guess, all parta the natchrl continyouem

What is clear rn is how plenny people spunkin' out the AI gonna reel folks in whatevah.

But how do YOU feel as an IMBIBAH?

Likely, before this all gits started wholesale, we can query gut/filosof Now.

So here's a question for evrywan:

Would I be happier 10 years from now if I knew my life depended more on infinitely efficient machines than potentially fallible humans?

In the age of information, ignorance is naht a choice.

I love Bard. Recently I used another AI Bot to give me a list on meditation mantras. Just a few minutes ago I pasted the same prompt in Bard. 2 seconds it blew the other AI Bot out of the water. Color me with a hot Damn for Bard.

Now I'm off to play with it some more.

That's great - people seem to be loving AI for many reasons... but keep in mind that AI content is not permitted when posting on the WF. From the rules:

Some of y'all may know this and some of y'all may not.

Y'all can use ChatGPT Playground to test out some different things with GPT. It lets you get more creative. It was put out by OpenAI...the same company that put out GPT.

Not sure why I'm saying y'all today.

https://platform.openai.com/playground

How does Bard and ChatGPT compare with Perplexity AI?

https://www.perplexity.ai


(I have no connection to it.)



I've given it a try with a few questions and it seems pretty good, i.e.
https://www.perplexity.ai/search/718...9a4542e9dc?s=c


I didn't have to log-in either, which was nice.

What are the differences between Bard and Chat GPT in terms of AI tool prices and plans?

Chat gpt still better, you can just input new data before asking it to make article. to ensure data is valid and up to date

In the meantime, I am awaiting the release of Bard in my region, and I have been using ChatGPT for the time being.

Chatgpt is better than Bard because Chatgpt has huge dataset than google bard.

I'm wondering if anyone else has noticed the answers from GPT seem to be changing in the way it answers...

I know it says it's being trained, but it seems the training is from all over the world, and it's affecting the outcomes?

I was trying to work with it yesterday and it seems the responses were more foreign-speaking in the way it responds. Not foreign language but foreign in the way I would put sentences together...longer words and more professor-type stuff.

Could just be me...but it seems the training is taking on a worldwide tone of voice.

I like Gili AI, most people don't know about it but it's my favorite

It's always ChatGPT.

We found ChatGPT is better for now

Haven't used both, but my guess is that if you are using it for content, the Google search engine WILL NOT see the content as original when using Bard whereas if using a non-Google entity to create content, Google will have no clue whether it is original or not

The principal contrast among ChatGPT and Versifier is the information source. Minstrel constantly draws data from the web, so it has the most recent data. ChatGPT's sources end with 2021 information, so it is restricted on more current exploration and data. Minstrel has more information to assemble data progressively by getting to the most recent examination.

Thing is, if'n we gaht a contest says DESCRIBE THE DISADVANTAGES OF AI IN ACTUAL ENGLISH, likely hoomans winnin' out rn.

You wanna be persuaded othahwise?

c/o a debatin' stance gaht you nailed?

Likely that is when AI ceases to be a benevolent servant.

Plenty debates gowin' on rn regardin' slavery an' enablement.

As (yawn) always.

Prahblem this time is, we more easily obliteratbyool by virtyoo of our communal suck.

So why pool drivel when evrywan's hair gowin' up in flames?

What CTA button you gaht says NO?

HECK, O PRINCESS -- ARE YOU SOME KIND OF
FLOPOUT GIRL?

Yurps. Spurps. Bibbidy Burps.

Make sure you read and understand the terms very clearly . Bard makes it quite clear that it will store what you input and use it .

You can opt out of certain options but read and understand the terms

AI is getting smarter all the time so take care of your information , especially personal details

A study released by Stanford University a few days ago says ChatGPT is actually getting dumber...

. The ability to understand math has dropped by 95%.

. Coding skills have dropped by 42%

. A steep drop in common sense and reasoning skills

. ChatGPT has become "safer but less rational"

According to OpenAI a few days ago "nothing has changed"

Are they wanting more people to subscribe to GPT4?

Anyways, here's the report from Stanford:

https://arxiv.org/pdf/2307.09009.pdf...-tell-you-this

anything new is supposed to be good. BTW i am hearing about Bard for the first time. i am going to try it now.

Chat GPT is getting smarter and smarter nowadays. We use it to build web page content and creative copies. Or translate text