GPT - Explained!

Let's talk about GPT, GPT-2, GPT-3 and ChatGPT in 10 minutes
ABOUT ME
⭕ Subscribe: kzbin.info...
📚 Medium Blog: / dataemporium
💻 Github: github.com/ajhalthor
👔 LinkedIn: / ajay-halthor-477974bb
RESOURCES
[ 1🔎] GPT-3 Main Paper: arxiv.org/pdf/2005.14165.pdf
[2 🔎] GPT-2 Main Paper: d4mucfpksywv.cloudfront.net/b...
[3 🔎] GPT original paper: s3-us-west-2.amazonaws.com/op...
[4 🔎] A very Nice intuitive understanding of GPT-3 architecture: dugas.ch/artificial_curiosity...
PLAYLISTS FROM MY CHANNEL
⭕ ChatGPT Playlist of all other videos: • ChatGPT
⭕ Transformer Neural Networks: • Natural Language Proce...
⭕ Convolutional Neural Networks: • Convolution Neural Net...
⭕ The Math You Should Know : • The Math You Should Know
⭕ Probability Theory for Machine Learning: • Probability Theory for...
⭕ Coding Machine Learning: • Code Machine Learning
MATH COURSES (7 day free trial)
📕 Mathematics for Machine Learning: imp.i384100.net/MathML
📕 Calculus: imp.i384100.net/Calculus
📕 Statistics for Data Science: imp.i384100.net/AdvancedStati...
📕 Bayesian Statistics: imp.i384100.net/BayesianStati...
📕 Linear Algebra: imp.i384100.net/LinearAlgebra
📕 Probability: imp.i384100.net/Probability
OTHER RELATED COURSES (7 day free trial)
📕 ⭐ Deep Learning Specialization: imp.i384100.net/Deep-Learning
📕 Python for Everybody: imp.i384100.net/python
📕 MLOps Course: imp.i384100.net/MLOps
📕 Natural Language Processing (NLP): imp.i384100.net/NLP
📕 Machine Learning in Production: imp.i384100.net/MLProduction
📕 Data Science Specialization: imp.i384100.net/DataScience
📕 Tensorflow: imp.i384100.net/Tensorflow

[@CodeEmporium]

If you think I deserve it, please consider liking the video and subscribing for more content like this :)

[@srishtisingh1437]

Would you be interested in conducting a workshop on How ChatGPT works, Please let me know as quickly as you can and I will share my email id so that we may discuss further

[@CodeEmporium]

We can talk about this by emailing [e-mail]

[@Quantum_Nebula]

This channel is so underrated. The way he explains complex subjects is remarkable.

[@CodeEmporium]

Thanks so much for the kind words :)

[@adamromanpaul]

You deserve it. ❤

[@NicholasRenotte]

My lawd Ajay this is absolutely amazing! Just binged watched this at the gym. Your ability to distill something so hardcore into such a simple explanation is nuts. Bravo homie!

[@CodeEmporium]

Thanks a ton Nick. Can’t wait for your course on this!

[@NicholasRenotte]

@@CodeEmporium at the moment it’s a minute and 45 seconds of me making language puns.

[@fahnub]

The crossover I never expected. 🙌

[@nilslusch6625]

Loved the video, it really helped me. I had a problem with understanding how the jump from language modeling to other tasks works. You saying that it's a finetuning step helped

[@li-pingho1441]

Wow, this video was so informative and well-explained. I appreciated the clear and concise way the subject was covered. Great job!!!!!!

[@CodeEmporium]

Thanks for the super kind words! More like this to come :)

[@anirudh514]

Very well explained

[@paull923]

nice thx! like the explanation of self-supervised in particular! wasn't aware of that

[@CodeEmporium]

Thank you for watching !

[@PauloCosta-ns9bf]

Thank you so much !!!👍👏

[@eng_ajy5091]

Really nice video, can I simply use pre-training GPT or BERT after fine-tuning for video captioning, which one is better for this task?

[@wiener2063]

Great video! Where did you get the numbers indicating how many samples it takes a human to learn those specific tasks? I'm wondering because it surprises me how low they are considering it takes humans years to learn how to form their first sentence.

[@shauryai]

Great work!

[@CodeEmporium]

Thanks a ton :)

[@user-dc1gu8ti2t]

Amazing Explanation

[@CodeEmporium]

Thanks so much for watching !

[@UbaidullahMOmer]

Hi bro you make crazy content. I really like your content. I am from Pakistan

[@ProfessorLinux]

🔥🔥🔥

[@NeoShameMan]

It's only know I understood why you were talking about workable memory lol, a working memory you work on...

[@vtrandal]

It's absolutely astounding how transformers learn to reason. It's ChatGPT's ability to reason that sets it (and others like it) apart from anything done previously in Machine Learning and Artificial Intelligence. We are truly living in the dawning of the age of artificial intelligence.

[@yookoT]

Is it that the GPT models can tokenize the text into the character-level for the lowest level? I was impressed by how it generated those very rare words which that BERT can't.

[@CodeEmporium]

It tokenizes text into “Byte pair encodings” which are subwords (like “-ing”). I believe this is done to balance the inference time offered by word tokenization and small vocabulary size offered by character tokenization.

[@yookoT]

@@CodeEmporium But when I asked the GPT-3 or chatGPT about some biomedical questions (the keywords of answers are already within the question), they can generate the words like "L-CCG-I" or "PNU156804", which, in my opinion, can't be processed or generated in subword-level.

[@alexkonopatski429]

But why is GPT better than for example BLOOM? Isn’t the architecture almost the same? So what makes the difference? (And here with GPT I mean gpt3, what makes GPT3 so good in comparison to others, even GPT2)

[@MBAFuturis]

Bloom not possible to sign up.:)

[@alexkonopatski429]

@@MBAFuturis what do you mean by that?

[@shivampradhan6101]

Bro your videos are really good. But I am not able to get an order to watch things. This is the 5th Or 6th video on gpt from your channel. Can you make an ordered playlist for all yours video for deep learning? That will be very helpful for people who want to consume your content

[@CodeEmporium]

Hey Shivam. Thanks! If you check the description of this video, you’ll see the main playlists now. There are playlists for my ChatGPT videos and another playlist for Transformers. Hope this helps :)

[@tsantoshkumar3935]

5:36 What is meta learning

[@blochspin]

These videos are content free. Its press science.

[@theJellyjoker]

After having ChatGPT talk to ALICE and other chatbots, I am convinced that ChatGPT is just another chatbot and not as revolutionary as people are claiming. I say this because it's very limited in its ability respond and falls back on repeating statements. Also after talking to it directly, I found it to be very simplistic in it' ability to converse. I'm sure the coding and underlying technology is very impressive but what it's doing isn't much better than a nearly 30-year-old chatbot and thus unimpressive from the standpoint of what it's supposed to be. I know someone is going to be very offended by what I just said and all I can say is, go tell it to the chatbot.

[@edilgin622]

I feel like at its core it has just memorised the entire knowledge of the internet so whenever you ask it to perform a novel task it just fails badly. Nevertheless it is still mind bogglingly complex under the hood thats for sure.

[@meguellatiyounes8659]

i dont understand the generative part of gpt . no body talked about it

[@CodeEmporium]

Hopefully I can help :)

[@Bella-gj6wc]

So by watching a movie on you tube, that has GPT, we’re inadvertently teaching GPT to learn and be better? I think I’m out on that, I’m not going to make AI’s advancement easier by engaging with it. I want NOTHING to do with AI! I may be the minority but I’m okay wth that. Thanks for the video.

[@tilakrajchoubey5534]

Where does llama fall in all these? Is it GPT based, BERT based or is different altogether?