GPT-4 Makes Nvidia the BEST AI Company on Earth (NVDA Stock)

let me tell you why I think Nvidia will become the most important company in the world ever since I was a kid I've always been fascinated by the idea of artificial intelligence I remember playing video games with basic AI powered enemies and being amazed at how they seem to be able to think and react to my actions in real time as I got older I began to understand more about how AI worked and I realized that it was about so much more than just video games for example I learned about how Google uses AI to show you the best search results based on things like an article's relevance and its popularity and since starting this YouTube channel I've learned about how AI was being used everywhere from Finance to healthcare to analyze massive amounts of data and make more accurate predictions about the future and today I'm Blown Away by the potential of AI what we're seeing with chat GPT has the ability to transform every aspect of our lives in ways we can't even imagine yet and it's not just open AI Google just announced their workspace AI which adds AI powered features to docs and to Gmail features like generating summaries taking notes during meetings scanning your emails and automatically recommending replies and so on and Microsoft just announced copilot an AI that learns how you use all of Microsoft's apps and how you interact with your co-workers then it uses that information to help you do everything from writing your next email to building complex spreadsheets to putting together your next presentation for a specific audience so whether we're talking about enemies in video games or search engines or chat Bots and virtual assistants AI is clearly touching every part of our daily lives but before you get too hyped up I have a confession to make after I went to grad school and spent eight years at MIT I realized that a lot of things that I thought were AI weren't AI at all for example that AI in my favorite video game was actually just a bunch of logical rules that determined in enemy's Behavior if I got too close to the enemy it would attack if the enemy's Health got too low it would run away so what I thought was AI was really a bunch of rules-based choices that each enemy would make to feel real sometimes but that fell apart pretty fast other times and Google doesn't really use AI in search either instead it relies on something called search engine optimization or SEO so if you search for best pizza in New York City Google looks at different factors for each result things like a web Page's content and keywords its Authority and how often it's linked to from other websites and even your actual location to determine the pages that should be displayed at the top of your search results don't get me wrong Google search is incredibly valuable but it's powered by tests checks and formulas that filter the best results to the top of Google's very long list of results which isn't the same thing as Ai and that brings me to the big question about all of these awesome tools that we're seeing come out of openai and Microsoft and Google are these things really AI believe it or not the answer is yes chatgpt is powered by a large language model called gbt4 GPT stands for generative pre-trained Transformer these three words are actually the key to understanding why Nvidia might end up being the most important company in the world not Microsoft not Google and not even open AI let's walk through those words one by one starting with Transformers as you know Transformers are robots in Disguise and they typically come in two flavors Autobot and decepticon wait not that kind of Transformer a Transformer is a kind of neural network which is a type of machine learning model that's inspired by the structure and function of the human brain neural networks are designed to recognize patterns and data and make predictions and decisions based on those patterns you've definitely interacted with neural networks before even if you're not a huge nerd like I am here let me give you a few examples you know how apps on your smartphone can detect faces and objects in pictures and in videos object recognition is done by a convolutional neural network or CNN I spent a lot of time on object recognition during my time at MIT you know how your smartphone's keyboard can auto complete your sentences and even gets better at it as you keep using it processing and predicting sequential data like words in a sentence is done by recurrent neural networks or rnns you know how social media filters can modify and enhance photos and videos in real time generating brand new images and videos that mimic real world data is done by generative adversarial networks or Gans Gans that's the kind of neural network behind open ai's Dolly image generator Snapchat and Tick Tock filters and so on one of the newest types of neural networks is called the Transformer the Transformer was actually introduced back in 2017 by researchers at Google to address some of the biggest limitations that the other neural networks I just mentioned have specifically when it comes to natural language processing the special thing a Transformer does is decide the relative importance of each word in a sentence and then use the most important words to guide its output that makes it really good at tasks like language translation text summarization Asian and language generation okay so Transformer is the T in GPT pre-trained is the p and it just means that the model has been trained on a large amount of text Data before being fine-tuned for a specific task like answering technical support questions or coming up with business ideas and so on gpt4 was pre-trained on a massive collection of websites books scientific papers and news articles with a wide range of topics and writing styles everything from Wikipedia to science fiction novels were included in the data set the use of pre-training in GPT allows the model to develop a general understanding of language that can then be applied to a wide range of tasks which makes it much more efficient and cost effective to develop more fine-tuned models for things like writing copy for websites making lectures for Education dialogue for movies and video games and so on so pre-training is the p in GPT now let's put it all together the G is for generative because the model generates new text it's hard to distinguish the text that gpt4 generates from text generated by humans because the Transformers in gpt4 were pre-trained on text written by humans now I know what you're thinking that's great and all but if Transformers were created by Google and then gbt4 was built by open Ai and then Microsoft is putting it in all of its apps where does Nvidia come in the answer is that all of these models are trained on nvidia's gpus by the way nvidia's GTC is right around the corner GTC is nvidia's developer conference where they showcase the latest breakthroughs in everything from video games and visual effects to Robotics and self-driving cars and every other industry where artificial intelligence is making a massive impact but this GTC is special because it focuses on generative AI like the kind found in chat GPT as well as nvidia's massive Omniverse program accelerated Computing and deep learning the exact things that I've been focusing on on this Channel GTC is happening online from March 20th to the 23rd and Jensen's keynote presentation is on March 21st and there will be exclusive sessions with a co-founder and chief scientist of openai the founder and CEO of Google's deepmind and tons of other awesome speakers and best of all anyone who registers for at least one session using my link can win an Nvidia RTX 4080 Founders Edition graphics card this thing is heavy all you have to do is attend at least one GTC session take a screenshot and send it to me after the conference by going to Ticker symbolu.com GTC I'll announce the winner for the giveaway a few days after GTC to give everyone who attended some time to enter nvidia's RTX 4080 is an absolute monster of a graphics card when it comes to high-end gaming content creation and professional visualization but it would actually take over 500 years to train gpt4 on one of those cards so what actually happened was that openai reached out to my Microsoft over five years ago to build an AI infrastructure that's capable of training their GPT model and Microsoft turned around and spent hundreds of millions of dollars developing that infrastructure while a lot of that money actually went to Nvidia to buy thousands of Enterprise grade gpus like their a100s and h100s nvidia's a100 gpus are designed for the most demanding AI workloads in data centers including machine learning deep learning and data analytics applications the a100 also features Advanced tensor cores which can do the complex Matrix math involved in deep learning much faster than traditional CPUs nvidia's newer h100 data center chips are a massive step up from their a100s in terms of performance promising up to a 9x speed up in AI training and a whopping 30X speed up for inference models like gpt4 I bet there's a ton of awesome information in Jensen's keynote GTC presentation and I'm super excited to break it down for you as soon as I can one one thing I'm sure he'll talk about is how nvidia's chips are designed to connect together to form more powerful systems at virtually any scale for example eight of these h100 chips can be connected to form a dgx h100 server system and if you combine nine dgx h100 service systems you get a dgx pod which is the reference design that can be scaled up to offer AI services at the Enterprise level and Beyond but it doesn't stop there if you think 32 dgx pods together you create a super pod which is one of the most powerful Computing systems in the world this system can handle massive amounts of data and can run the most complex AI workloads with ease and Nvidia is linking 18 of these super pods together to build a super computer called EOS which will be about two and a half times more powerful than fugaku in Japan which is currently the world's most powerful supercomputer and this isn't just a flex every cloud computing platform from Microsoft Azure to Amazon web services is going to need the these enterprise-grade gpus and many of them in the next few years when chat GPT came out a few months ago it was using gpt3 the gpt3 model has over 175 billion parameters in it at a high level a parameter in a machine learning model is a little bit like a variable in an equation so 175 billion parameters is a very complex model in fact gpt3 was the biggest machine learning model by far when it came out well gpt4 has about 10 trillion parameters which means the model is almost 60 times bigger than gpt3 that's a massive model to pre-train that P in GPT and that doesn't even get into all the ways that other people will want to train the final model to do things like write copy for ads and websites dialogue for films and video games Bedtime Stories and epic novels and so on here's the thing that's just text generative AI can also make images and Graphics think about designing logos with no graphic design experience generating realistic product images before ever building the product fashion design heck you can even have ai generate the models wearing the clothes not just the clothes themselves the people modeling the clothing in these pictures don't exist they were generated by AI Nvidia canvas is an AI driven program that helps people turn simple shapes and lines into photorealistic landscapes in near real time the program has different brushes for materials like mountains and rocks rivers and oceans fields and flowers and all sorts of skies and clouds it also lets you change the lighting and the time of day think about how useful this is for concept artists or storyboarding scenes and movies and in video games or just touching up your own favorite photos without needing to learn a new tool or how about generating 3D models and assets for video games or architecture or product design all of the sudden something that took serious industry experience can be done from a set of prompts it's not just about making models without knowing these tools it's about AI being able to generate models much faster than humans can which means much lower costs and much more room for iteration or the AI could create 3D models that human designers didn't even have the imagination to consider then they can import those assets to nvidia's Omniverse to see how the 3D model behaves in real time and make changes before creating the physical version so an architect can create a 3D model of a house and simulate how natural light would enter the building at different times of day Omniverse can be used to simulate real world scenarios for example a car manufacturer can simulate how long it would take to make a car given different Factory configurations and then choose the best one it's only a matter of time before every asset we just talked about from the chairs in a house to the house itself to large factories filled with conveyor belts and robots to entire virtual worlds are generated by AI instead of being made by human hands and it's all powered by thousands and thousands of Nvidia gpus and that's why I think Nvidia will become the most important company in the world so don't forget to use my link in the description below to sign up for at least one session for GTC seriously the sessions range from learning about how to use generative AI at your own job to hearing about the breakthroughs in self-driving Robotics and everything in between but there's one more disruption that we really can't ignore so make sure to check out this episode next and if you feel I've earned it consider hitting the like button and subscribing to the channel that lets me know to put out more episodes like this either way thanks for watching and until next time this is ticker symbol you my name is Alex reminding you that the best investment you can make is in you

Channel: Ticker Symbol: YOU

Views: 90,671

Keywords: ARK Invest, Cathie Wood, arkk, arkk etf, growth stocks, tech stocks, stocks, ticker symbol you, inflation, fed interest rates, tsla, arkk stock analysis, should i buy arkk, best stocks to buy now, cnbc, nvda stock, nvidia stock, msft, msft stock, microsoft stock, chatgpt, openai, msft openai, best AI stocks, openai stock, invest in openai, tsmc, tsm stock, GTC 2023, Nvidia GTC 2023, GTC Keynote 2023, Jensen Huang, GPT-4, gpt4

Id: wHPwpq0Dy8s

Length: 13min 47sec (827 seconds)

Published: Sun Mar 19 2023