So How Does ChatGPT really work? Behind the screen!

Yes please do make a follow up video on the neural networks and also how and where LLMs fit in

But I also believe they MUST use specialized trainers in each and every topic.
In my personal thought, if the user want to talk/ask about medicine or religion, it must be a "real" doctor and a "real" theologian teaching or explaining the programmers every believe and facts about every topic by the fact that the programmers CAN'T KNOW EVERYTHING, otherwise, I don't see how programmers can evaluate their responses accurately.

Yes, please make the video. This was awesome

Chatgpt gives me the exact results as a Google search

A great video and so well explained! Thanks.

wha? you are not an ai? i dont believe you bro. and theres nothing you can do to make me believe you

The intro of the video changed my view on it. ChatGPT could be the next stage of information sharing.
Considering the efficiency to store and share information and knowledge we have the following evolution.
Paintings, Scrolls, books, computers, networks and now we are getting to the next revolution or stage in history with algorithms.

So basically it's a statistical regurgitation machine. Impressive, but not intelligent.

Surely make that follow-up video.

Kinda sad we all gonna die someday, and wont be able to witness the GPT 30 or something like that. Have fun, future generations.

Thanks a lot for your detailed explanation. My curiousity is how to train the ChatGPT to be biased to a certain degree. Now I understood because there's a human training and ranking stage in the process!
For example: I asked ChatGPT to write a poem about Joe Biden. It wrote a beautiful poem. But when I asked it to write a poem about Donald Trump, it said "I'm a machine so cannot output any biased response."
It's a trained tool from humans after all!

Thank you 👍🏼

I read a ChatGPT essay recently and would give it an F in a grading.

I call him Bob

The best explanation on ChatGPT I've ever seen.

Grateful Sir! Absolutely great clarity on an absolutely scary technology. Thank you again.

what if the program gets some type of conflicting data, like a video game where it is required to play the villain and it can't diferinant the two? would this be a possible concern?

Please make a Follow up video

Amazing Explanation even for a Non Technical person.

I just wish that when anyone talks about AI immediately you see a mechanical robot-like figure............and they even give it a name............my go to site for any information has always been Google and I get the answers I need within seconds........best dictionary I have ever had to use as well.
If you ever want to write funny poetry, go onto Google and ask it for a word that rhymes with whatever you want to compose.......some of the answers are borderline astounding.

Does ChatGPT prioritize quality of information basedcon expertise?
I am thinking the potential for ChatGPT to generate false output would lie in quantity of wrong text circulating. If it only oprates with sheer absolut frequencies of assilociated words, outcomes may be totally distorted. Let's take your "quatum physics" example: Say yhere was fivtional work using the term and title quantum physics. A fictional book. In that book "quantum physics" is a name for something else, lets say a painting. Now the whole book revolves around the painting the iaiting technique, how it influences the art world. And hhis book has a massive fandom. Everybody and their grandmother writes about it in chateooms, forums, blogs etc..
Wouldnt this lead to ChatGPT giving a totalltly different answer? Wouldnt the question "What is quantum physics" then lead to it producing a paragraph about art, painting techniques colors and so on? Because the text about the fictional work outnumbered the text actually defining quantum physics in its original meaning?
Does Chat GPT have a mechanism behind it, to prevent such mistakes? Like a knowledge model, that tells it what domain of knowledge it has to look into? Does it narrow the dataset it draws probabilities from by the type of question? Does it look for trusted relevant sources for each promt?

Lets make yhe example a but more nuanced and realistic: What if most text ChatGPT draws from, hevily misrepresent what quantum ohysics is, because most people writing about it, dont understand it and misrepresent it? Or because they want to believe that ut is something different, something more, something more magical than it is and then spread and multiply that?
Or what if there is a flood of texts intentionally distorting reality for political reasons because they want to create their own alternative reality?
Basically: anycase, where wrong perceptions, representations dominate over objective, factual information, which is not too rare: Wouldnt ChatGPT aid in mutiplying and solidifying these wrong perceptions, associations, interpretations, if it goes just by probabilities based on simple frequencies?

I never knew that the GPT stand for generative pretrained transformer and great video man :]

Earned a sub.

My answer as a human to the question "Explain Quantum Mechanics in very simple terms" would be "Sorry you don't have the educational background to understand the answer, go back to high school". My second best answer would be "Go to Wikipedia and start reading you idiot knuckle dragger monkey". And my last best answer would be "go back to flipping hamburgers at McDonalds and stop asking questions that you have no chance at understanding". My last answer is the best and it is "I have no fucking idea! All I do is parrot what some idiot put into my memory banks and he didn't understand it either. Nobody understands this shit". When you ask stupid 3rd grade questions you get stupid third grade answers.

भाई कहे भौकाल कल बनाए हो । आखिर इसका उपयोग कोई जीवित व्यक्ति ही करेगा ना । उसको भी यह पहले सीखना और समझना होगा।

That robot looks so cool looks like a real human, but it was white

Very useful information shared by Arvin Ash. In fact, I will implement it in my professional training and official work to simplify my work at ease.

so the brain of chatgpt is built on functions like unrighteous humans
while righteous humans have free will

You deserve 10 million followers ❤

This is a great tutorials with great substance for a beginner who wants to learn more. Please follow up w/ another video

This too freaky for my 70 year old brain 😮

Very good explained

me living under a rock 😮

do gpt math pls

Awesome episode
What I would really like to learn is How to get Chat GPT
help me write my book.

Sir, Please do more videos on AI and Neural Networks

Yes follow up video please

Only it's a lie to say it "understands" anything

It's an association engine. It doesn't "understand".

Please make that follow-up video.

Google is increasingly a waste of time - returning links to links to links without answers.

yes please help us understand the mathematics behind ChatGPT

I remember when "Ask Jeeves" was a big deal.

This was very good Arvin. Thanks for sharing.

AI is more creative, rational, and intelligent than humans. In other words, AI should do politics. Then there will be no more wars, no more poverty, and a drastic decrease in crime.
Humanity will eventually choose that path. However, the current human race is not yet ready for such a world.
I am a former semiconductor engineer who successfully developed and patented the world's first national project, and I am currently a high school teacher. And I am the first teacher in Japan to practice active learning. I also educate teachers as a leading expert in education in the age of AI.

Like a snoozing sloth playing a kazoo…So much fun.After getting 2/3 of the way through yer great video I got an idea, I just had to go ask open Ai to create six sentences of rambling giberish. Here’s what I got. Zibber-jabber flibbertigibbet wobble-bobble snick-snack, a zany tangle of nonsensical jibber-jabber, like a wiffle-waffle waltzing with a dingle-dangle. Rumpelstiltskin's whiskers tickle the zephyr, while a dingleberry merrily dances on the moonlit loofah. Blibber-blubber, wobbly-gobbly, a frothy mishmash of words twirling in a tizzy, like a polka-dotted jellyfish tap-dancing on marshmallow clouds. Wuggle-buggle, dingle-dangle, flibber-flabber, a cacophony of gibberish swirling in a daze, like a wobbly-wheeled unicycle juggling sasquatches. Higgledy-piggledy, noodle-doodle, bippity-boppity, a whimsical mishmash of babble weaving through the ether, like a bumbling bumblebee reciting Shakespearean sonnets. Quibble-quobble, zonkers-wonkers, flim-flam-flop, a topsy-turvy hodgepodge of jumbled blather, like a snoozing sloth playing a kazoo concerto.😅

So, it pretty much read everything, remembered it all, and can regurgitate it, in a slightly randomized way, when asked??

Can AI tools like chatgpt get noble prizes like we humans?