Large Language Models: Part 2

How do language models like GPT and Palm work?
Part 1: • Large Language Models ...
See next: text-to-image (Parti, Imagen, Dall-E): • Text to Image in 5 min...
0:00 - intro
0:14 - next word prediction
0:20 - word embeddings
1:01 - transformers
3:11 - generating text
4:13 - stacking attention layers
4:47 - training data
5:21 - GPT-3 examples

[@fazlfazl2346]

The good thing with short tutorials is that if they are good, you will save hours of work. If they are bad, no time is wasted. This is a good one. Saved hours of work which I would never do anyway.

[@5133937]

This is one of the best explainer vids on LLMs I’ve seen yet. Not too long, not too short, good pacing, good visualizations. Great work, thanks!

[@kahoku451]

It’s amazing and so funny to me how LLM can produce fully functional python programs and write these poetic Bob Dylan inspired lyrics… but when prompted with “you’re going north. You turn right, then you turn left. Now you’re going …” it said SOUTH ☠️

[@samandoria]

Just like human "brain cramps" or "brain farts". The problem is that the current models aren't self learning or made to analyse their own answers so they don't correct these error before output or can correct the weights in the future when confronted with new data.

[@TimHulse]

It Re-raises old but interesting points about the difference between language and logic. You can make linguistically sensible statements that defy logic. Try the book “Godel, Escher, Bach” if you like this area.

[@webcache1000]

6:02 If you are just just south of the North Pole, turn right and turn left over the North Pole, you would be heading South. I suspect it is getting confused due to the similar riddle where if you head North, turn right and you are now heading South - where are you? It has probably seen this riddle a confusing number of times and weaved that into its weights/response.

[@distrologic2925]

Because it doesn't necessarily understand how directions work, it only knows how "north" and "right and left" has been used in language before, and is only an estimation. People are unlikely to talk about celestial directions and turn right or left, so there was probably not enough similar to your query in its training data that it could draw on and other examples such as "north and south" have taken more weight.

[@swanandjoshi333]

Logic is a whole different thing ig

[@TimHulse]

I really enjoyed both of these LLM videos. They are so concise and informative and the pacing is excellent.

[@whannabi]

He is an Arab from the Middle East. My dog wants a walk.

[@apc067]

I [or shall I say, my internal GPT?] first misread your comment as "... and the pancake is excellent". 😃

[@csmac3144a]

This should be adopted by high schools everywhere. Superb teaching. A bit fast for me (I’m 61), but I can re-watch until I get it.

[@f18a]

Great stuff. Possibly the best intro material to LLMs that I have seen. Thunbs up!

[@Charlieee1]

I like that you used a recipe prompt to demonstrate what a LLM is good at doing, then actually followed the recipe and proved that it actually worked (and tasted good!).

[@markbordelon1601]

part one and this part two are masterful introductory teaching

[@roywem]

This is the single best explaination I’ve come across on LLM’s

[@stuartthomas7441]

So pleased to get a clear and credible glimpse under the hood. Thank you.

[@scottlott3794]

Dude, I have been researching Transformers and Attention for weeks. You’re a master mind, I get it now!

[@ianassarandas]

you actually made the cookies hahaha that's awesome, great video btw!

[@poklet]

Thank you. Those were very clear explanations of just the right length. Loved that you cooked the pancakes too!

[@yojimbomk]

Just 16 minutes, it is amazing. These are the best concise videos on the topic including basics in neural networks I have seen. Many thanks for sharing your knowledge!

[@shoe_Bin]

Soothing voice to learn about transformers and warm room is perfect for recipe for my sleep.🛌

[@aidarfaizrakhmanov1901]

This is a really good intro indeed! I encourage to make more content like this

[@MaleGeminiCat]

These are two great videos that introduced how large language model works in a very comprehensive way.👍👍👍

[@janicem5942]

Brilliant overview for a non-technical person like me... and glad to see you tested the recipe!

[@jaunalapa]

You, sir, have talent for teaching. Even with my fairly technical background and watching a lot of videos I was struggling to get my head around LLMs until I watched your videos. Hope you continue what you started it would help a lot of people especially now when we see a lot of people trying to run their own local LLM versions on their PCs.

[@somethingness]

Typo alert! Frost's wonderful poem begins as so: "Two roads diverged in a yellow wood". (Not "diverted"!)

[@vast634]

You really have a talent to teach things.

[@eenvleugjegoeiegames]

Man you NEED more subscribers, the content and video quality is way too good for only 15k!

[@EGlobalKnowledge]

An excellent video on language models

[@fernwood]

This brings me one step closer to understanding. Thanks.

[@ypetkar]

Very informative and concise videos to understand LLMs their complexities and what i takes to make these models successful.

[@0815Snickersboy]

This is the best video on Language Models that I seen. Probably the best on the Internet. You should maybe add chatgpt to the title to get more views.

[@smudgepost]

Very much enjoyed these two videos. More please! Clear and detailed.

[@sreenivasulupala2374]

Excellent explanation. Thank you.

[@craftycurate]

Really easy to follow, well paced, easy on the ear, and just the right level thanks!

[@jteichma]

What a great walk through! Thanks so much for sharing.

[@xybnedasdd2930]

I think this is the best, most intuitive and most illustrative video describing LLMs/transformers. Thank you so much!

[@DavidDiaz-zg5sv]

I love that you finished cooking the recipe! Great video :)

[@ravinatarajan4894]

Thanks for the crisp walkthrough of the technology. It is a very good introduction.

[@Paul-rs4gd]

Thanks for an original presentation of Large Language Models. It gave me new insight.

[@KalebPeters99]

Not a second wasted. Just brilliant ❤️

[@roscatres]

I hope you keep up with these videos, they are seriously great. Already suscribed and I'll check the rest of your channel. Thank you.

[@DrNuyenVanFaulk]

Thanks so much for both of these videos. They are wonderful. I think I understood them a bit more since I’ve done some basic assisted machine learning dev (up to neural networks). If anyone is a bit lost, read up on linear and logistic regression, then onto neural networks.

[@sciexp]

Very interesting... Now I understand more about how ChatGPT works...

[@MindyMcAdams]

Thank you! This is excellent. I love the animations. They are helpful!

[@KumquatChampion]

This is awesome. Would love to learn more!

[@GodofStories]

Absolute best explainer. Where have you been though since GPT-4?? Would love to see you do more content. You have a talent!! I think you might be busy working on LLM's for research or in a company, but would be cool to see more videos!

[@stefanbiesdorf4637]

Thank you Steve for a personal moment of enlightening by opening the "black box of AI". This is an outstanding educational piece, in particular in combination with the first part.

[@saisaske1]

Very good explanation

[@frank-reneschafer5512]

Great effort - nice video, good and comprehensive explanation.

[@ayushsharma9036]

That language neural network at 0:50 belongs on a tshirt somewhere

[@tvillaluz]

Haha you made the pancake

[@bloodywolftr]

Pretty nice explanation in both videos.Thanks!

[@AncientSlugThrower]

Great video, leaving a comment to let you know it was very insightful. Thank you.

[@AMGbot]

Loved this! Thanks for the great video!

[@TheCaioKyleBraga]

This is my first search for a LLM explanation and very pleased with the video. I am not a mathematician or programmer but I am very interested in learning how LLM works. From my humble perspective I can say we reached a point of no return and this technology is progressing at an exponential rate. With the development of quantum computing, I have no doubt that it will surpass human intelligence in ways we don't understand.

[@dv6165]

This was awesome! Thanks

[@StratosFair]

Great explanation, glad I stumbled upon it ! Subscribed :)

[@bingolio]

EXCELLENT Vid, PLS DO MORE, on Deep learning , covering the whole workflow of making an LLM, especially, what os LORa, VEctor embeddings, etc I'm sure you'll get huge interest You have a gift for explaining. Thanks!

[@preethamrangaswamy7371]

Very underrated and underappreciated video.

[@danjsy]

Great, thanks so much, massively useful.

[@coraltown1]

the Mona Lisa of LLM explanations .. thanks!

[@g5min]

Okay, that's the best comment yet -- thank you :-)

[@user-hf5og1bt8j]

amazing video! thank you!

[@padetiit7014]

I loved that you actually cooked that recipe! :-)

[@Patapom3]

Very well made!

[@jorgesoberon6866]

Very clear. Thanks a lot.

[@Sukant98]

amazing videos!! learnt so much

[@fenandamilanda2032]

nice work, really good to visualize these things even though I already know this.

[@yourfuneral]

Thanks for the good explanation, very much on time

[@ivanocj]

Very good content! Keep going! thanks.

[@christianwestermann4680]

Well done!

[@xflory26x]

Please do more videos on LLMs!!! But also I need to know, how were the pancakes?

[@carrumar]

Great videos! Btw, how did you like your avocado cocoa thing? 😄

[@rajachan8588]

Awesome. Thank you

[@thomasforrest1931]

great video bud cheers!

[@prathams8685]

Big thanks mate

[@ChatGPt2001]

Continuing from the previous response, here are some additional considerations and challenges when working with large language models: 14. Data Privacy and Ethical Concerns: - Be aware of privacy concerns when collecting and using data for training. Ensure that you have the necessary permissions and comply with data protection regulations. Ethical considerations, such as bias and fairness in the data, also need to be addressed. 15. Compute Resources: - Training large language models requires substantial computational resources, including high-end GPUs or TPUs and large-scale distributed computing infrastructure. These resources can be expensive and may not be accessible to everyone. 16. Energy Consumption: - Training large language models consumes a significant amount of electricity, contributing to environmental concerns. Some organizations are actively working on making AI training more energy-efficient. 17. Model Size and Efficiency: - While larger models tend to perform better, they also require more memory and computational power. Balancing model size and efficiency is crucial for real-world applications, as very large models might not be practical for all use cases. 18. Fine-Tuning and Transfer Learning: - Fine-tuning pre-trained models on specific tasks is a common practice, as it requires less data and computational resources compared to training from scratch. Understanding how to effectively fine-tune models is essential. 19. Evaluation Metrics: - Choosing appropriate evaluation metrics is critical. Different NLP tasks may require different metrics. For instance, accuracy may be suitable for classification tasks, while BLEU scores are used for machine translation. Select metrics that align with your objectives. 20. Bias and Fairness: - Large language models can inherit biases present in the training data. Mitigating bias and ensuring fairness in AI systems is a significant challenge. It requires careful curation of training data and ongoing monitoring. 21. Robustness and Safety: - Ensuring that large language models are robust and safe is essential. This includes protecting against adversarial attacks, avoiding harmful or inappropriate outputs, and ensuring that the model behaves predictably. 22. Research and Collaboration: - The field of large language models is rapidly evolving. Staying up-to-date with the latest research and collaborating with the AI community can help improve your understanding and the quality of your models. 23. Resource Sharing: - Due to the resource-intensive nature of training, sharing pre-trained models, datasets, and trained weights is common in the AI community. Leveraging existing resources can save time and resources. 24. Ethical Considerations: - Consider the ethical implications of your work. The power of large language models also comes with responsibility. Engage in ethical discussions and follow guidelines for responsible AI development. 25. Interpretability: - Large language models are often criticized for their lack of interpretability. Efforts to make AI models more understandable and explainable are ongoing to build trust and ensure accountability. Training large language models from scratch is a challenging and resource-intensive endeavor, and it may not be necessary for many practical applications. Leveraging pre-trained models and fine-tuning them for specific tasks is a more common and efficient approach in many cases. Moreover, ensuring responsible and ethical AI practices is paramount when working with such powerful language models.

[@koho]

Great overview! People need to see this video pair before freaking out that LLMs are actually intelligent.

[@moedemama]

Cool video, thanks

[@abhaychandrol]

Thanks

[@andresroca9736]

Great the concepts transition. Great illustrations. The best of the best this couple of videos. What about more on other networks like r-cnn and audio nets? 😃

[@g5min]

Thanks! I'm working on one on reinforcement learning now...

[@bgustinjr]

Thumbs-up for actually making the pancakes. 😂

[@deand6411]

Thank you. Just the right level for my tiny organic brain.

[@food4yann]

I love the fact that you really tried that recipe! How was it?

[@CoreDump07]

liked and subscribed

[@mineralt]

those pancakes look pretty good

[@DJWESG1]

Ever think about audio synthesis and wave forms?? And how analogue synthesis utilising wave tables can offer a way to both communicate and comput information.

[@SethWieder]

I would love more of the visual explainers on ML concepts. Subscribing. "Some folks say they're overhyped / But I do think that's true / I think they're just misunderstood / Just like me and you" That generated lyric gave me CHILLS

[@kahoku451]

Sameee

[@lunaxquinn]

The important question though is, were the pancakes good?

[@richardharris9708]

ChatGPT gets the 37 question right now.

[@leeamraa]

fun funny fantastic and I am a fan!

[@dormin1850]

This was super informative and concise, loved it! But my real question is how were those pancakes?

[@KeikosCake]

pancake approved 🥞👍

[@chnolte]

Just look how far we have come in only eight months.

[@triton62674]

Fantastic work, hope those pancakes tasted better than they looked! xD

[@g5min]

they were seriously delicious

[@Ai.Sentinel]

Was the chocolate guacamole pancake any good?

[@zuqini]

I'm a bit confused by how stacking attention layers works at 4:12. Does the second layer take the first layer's prediction as input? Is the first layer's prediction still "next words" at that point, or is it now some sort of abstract intermediate value? How exactly does that capture higher level reasoning? Would appreciate any clarification!

[@martinstu8400]

eliminating bias and stereotypes from language models is a lost cause, because it's the same as asking the network to lie.

[@parabolicpanorama]

you didn't understand anything did you

[@speicaldark]

This is really nice! Now I understand why chatGPT tends to make up a lot of stuff with coherent sentences

[@whannabi]

What I'm wondering is how are they correcting its errors. For traditional NN, we have heat maps but I'd like to see something similar with transformers at the highest level to see what kind of patterns it noticed. Maybe that's what they use to correct its mistakes

[@gideonk123]

@@whannabi ChatGPT used a process called reinforcement learning from human feedback (RLHF): They used an already trained GPT-3 which already at the time. Humans both submitted new sentences as input prompts to ChatGPT being trained, and also ranked the output (responses) of the model. Then the ranking of the responses were then used as reward targets to continue training the model to obtain more desirable responses (measured by how the response rankings had increased).

[@DonaldTamMisterDee]

This is super fascinating. I want to learn to grow my own language models from the ground up with languages like c++/rust. I dont care if the language model i develop is inaccurate or not. I just want to understand chatgpt under the hood. How may i get started?

[@spenarkley]

Why dont you ask the language model how to make a language model?

[@infiniteplanes5775]

I have! Really great tips!

[@parabolicpanorama]

ask gpt to stack transformer layers and add some input output layers. add some loss functions to optimize and get a large dataset.

[@mlusalin2379]

It looks like GPT hands you a great chocolate guacamole pancake recipe... now I wanted to try too loool

[@genegade]

Verdict on the guacookies?

[@efisgpr]

GPT-4: Might I suggest dubbing them "Chocomole Pancakes"? 😮😂

[@GeorgeJohnsonJackofAllTrades]

Rather than asking chatgpt to translate, I recommend asking the question in the target language. You'll significantly different results - more accurate in the target language. Stated without proof or explanation which should be intuitively obvious after watching these two videos...

[@altalt7653]

I can't get why we stack them like so. If the first transformer block predicts a word, what second does, third? And why do they still need attention then?

[@jaredf6205]

What do you think of GPT4?