Why You Should Learn About Generative AI, LLM's Models In 2024?

hello all my name is kushak and welcome to my YouTube channel so guys in this video we are going to discuss about why you should know about generative Ai and llm models uh because recently I have got a lot of questions from students from professionals like how we can use generative AI in their business use cases in their companies and all right now see uh and recently i' also taken a live webinar to make you understand what is exactly uh generative AI what is the difference between machine learning and deep learning what is discriminative models what is generative models you know so considering this I will try to answer this particular question why you should definitely know generative AI in 2024 and you should also know about llm models also right now from the past 5 years guys I have been uploading a lot of videos on machine learning deep learning probably I've shown you even Eda feature engineering I've show you deployment techniques I've shown you mlops techniques I've shown you creating end to end projects many more things has actually happened right now what were the major challenges that we were facing when we were probably working as a data scientist where we are using traditional machine learning algorithms and deep learning algorithms try to understand this specific thing very much clearly then you'll understand the importance of generative AI so when I was probably working from 2014 okay let's say I'll give you an example from 2014 and probably I've been uploading videos YouTube channel from 2018 itself okay now during that point of time whatever problem statement we used to start with right let's say I want I want to probably create an ml project for a specific business use cases how that entire life cycle used to go ahead with probably start with data collection data collection was a difficult challenge in any companies you know because to start any machine learning or data science project you require data you require a huge amount of data please understand this video guys this is super important right why I'm specifically focusing you all to learn generative AI okay there is a reason okay I know you just tell let me know in the at the end of the video whether you agree with me or not but please make sure that you watch this video till the end and again the like Target for this video will keep it to thousand because I am going to tell you a lot many things over here right because this is what companies thinks you know and why they should harness the power of generative AI many more many more topics I'll be discussing in this specific video okay so please do hit like please do make sure that you share this video with everyone out there now initially when we used to do machine learning projects the first if I probably consider the life cycle of a data science project the first step was what was it it was data collection now data collection is a big challenge for any use cases yeah there will be some data with the company but again to solve a business use case accurately to provide you a good accurate result you know you really need to collect data from different different sources and after that the raw data once you probably get it right you you let's say you're taking it from 10 to 15 different sources then the next challenge was that how we can probably make this specific data much more clean and that is where Eda feat enging many more steps were actually done right and then after you do all the steps then you create a model right you train you train this specific data set with your respective models whatever machine learning or deep learning models you are actually creating right now in all these steps trust me there were a lot of efforts that was put if I talk about feature engineering right there's a very good saying in the data World 30 percentage of your time of the entire project just goes in Ed and feature engineering right now what is happening with the help of generative AI just imagine this see in generative AI when I talk about various l m models right large language models right now this models are already trained with huge data set right and when I talk about this kind of llm models most of the different different use cases it is automatically able to do it for you right it has those capabilities right let's say I want to do any specific NLP task in my business use cases there it is so how I will probably start start everything from scratch invest so much of time let's say 6 months I'm probably taking my entire time to probably build a specific project I have to put a lot of resources I have to use a lot of developers I have to make sure that each and every stage I have to make sure I do it in a proper way and later on if I don't get a model which is of good accuracy then again I have to repeat this entire process right but now most of the steps in this since I already have some models which are trained in huge amount of data which is able to give you a generic accuracy of good good because it is already trained with huge amount of data right and it is already giving you a good General accuracy right now just try to understand in this if you are just able to fine tune with your data set on those specific model right and later on what you do in this model once you finetune it you can integrate with any other application now this way it is reducing the development efforts see one important thing is that some part of the life cycle of a data science project is being already taken care of but there are steps there are steps there are some mlops activities that you have to do that you have to do it for every project Dockers kubernetes cicd pipeline everything you really need to do it so I'm saying you guys in 2024 there are two important things that I'm really going to focus in my channel I have uploaded videos on machine learning I've uploaded videos on stats I've uploaded videos on feature engineering I've uploaded videos on machine learning I've uploaded videos on deep learning and this year also I have uploaded many endend projects where I've where I've shown you the power of mlops tools right and with all those things the same thing is basically getting used in Industries in the real world problem statements so in 2024 again I'm going to focus on generative Ai and mlops see mlops is something that is required because your developer needs to make sure that the entire project process right from QA to u to production how it is specifically going how each and every experiment is tracked how you are able to probably add new data set and do all this retraining of the model regularly everything of this particular process is automated by this mlops activities right now let it be generative AI let it be any data science project now what exactly generative AI is able to do it to you it is automating most of of the life cycle of a data science project because you don't have to even worry about all those things most of the thing you have to worry about how to do fine tuning and once you probably get the model how you can deploy that model efficiently and that is where these big companies let's say you're talking about paid llm models right recently one of my student were asking Krish my manager told me to use open source llm models right so is it possible can I do it or directly should I use open AI llm model see openi LM models when they actually provide you Let It Be GPD 3.5 GPD 4 gp4 Turbo anything they also provide you an API through which you can also fine tune in their specific Cloud space itself right they have Cloud space they have that entire infrastructure so whatever charges they are basically charging they are charging based on the API request how many number of API requests based on the token size right if you also want to go with open- Source llm models you can definitely go it right you can download all the open source llm models on top of that you can fine tune with your data set but the major important thing will be that when you're using all the mlops activities when you deploy this application because the real cost real cost will be in fine-tuning because you have to use gpus for fine tuning purpose and the second cost will be inferencing from those llm models what does inferencing mean you're sending a request to that llm model you're getting some response back so for that you need to do the deployment let it be in AWS there you need to enable gpus over there so that you get the response very much quickly right now you are using open a API right you ask any question it'll be able to give you quick response and for that only they are specifically charging but in the back end they have huge Cloud spaces huge Cloud setup GPU setup everything so from fine-tuning to inferencing everything is basically provided by open AI there are also some other platforms I've heard about some platform like bradi AI okay now what does gradient AI do is that they provide you three open source model right one is Lama I think they also provide you uh mrol I think it is a paid one and they provide you Cloud space and GPU setup to retrain or fine tune your data based on that you can create your own custom llm models and after that they'll also give you an API so again there is a cost that will be involved based on the API request that you're sending now companies just imagine they were taking so much of time in spending in probably setting up each and everything right from Cloud to developers to mlops everything it was taking time but now because of this companies like open AI or Google or Microsoft or any models that are probably coming up in this uh space the llm models they are also trying to provide you the cloud setup or the GPU space that is actually required so that you don't have to worry about the infrastructure part and that way you able to create your llm application now let's say there are some startups they think that no I don't want to probably put a lot of money of mine in this kind of infrastructure itself right they can directly use that but there are also startups who will be thinking why not just take the llm model fine tune it and we will try to deploy this LL model with our own infrastructure in the cloud that is also possible Right understand this two specific thing and that is the reason guys you should know about generative Ai and llm models because it is making your task very much easy right now at the end of the day people may ask CHR then what is the use of machine learning deep learning why we should learn it understand one thing guys machine learning deep learning you know for any anything to be getting stronger right later on you'll be doing fine tuning you'll be doing some of the task right you require some basic blocks some good amount of basic knowledge and that is what you're getting from machine learning deep learning feature engineering and all that is the best way to learn tomorrow if you go for the interviews they'll not directly ask you about generative AI they may ask you about mlops they may ask you about deployment part they may ask you multiple things they may ask you some questions on generative a what all task you specifically done but most of the question will revolve around in asking about machine learning and deep learning this is super important to understand right so whether you are a developer whether you are any person tomorrow if a if in your company your manager says Hey implement this thing you should be able to give an idea yes I can actually do it and this is done by one of my student I told you right he his manager told that hey we want to quickly Implement a chat bot okay and he told that don't use paid tools okay use open source and then we will try to look at the infrastructure because we are able to Bear those cost I don't want this cost to go uh very high because if they are probably using cloud services or infrastructure from other company it may take more time right so what they wanted they wanted to use open source they wanted to do fine tuning and they want to take care of the infrastructure and he was able to do it and from that video of Lama 2 he was he he was able to do it very much easily right so that is what I'm actually saying right we have opportunities we can look for it and by this your task becomes really much easier and you get a good accuracy model right that is the most important thing and as we go ahead more good models are specifically going to come up right and it is being you just need to worry about the entire project itself sometime you need to worry about the infrastructure infrastructure in short and sometimes if the if you are okay to Bear the infrastructure cost from the company that is providing you you can also go with that but in short you able to create powerful use cases so my request for you all is that yes in 2024 generative AI looks super amazing super cool that is what I feel that is what is my gut feeling and this information the gut feeling that I actually bring up is by talking to many people through many people um who have already made this transition and working on this kind of projects right so I hope you like this particular video this was it for my side I'll see you all in the next video have a great day thank you one all bye-bye have a great day take care

Channel: Krish Naik

Views: 16,566

Keywords: yt:cc=on, generative ai tutorials, llm models, open source llm models, deep learning tutorials, google gemini pro, open ai gpt 5, artificial general intelligence, why to learn Generative AI

Id: IJeaVHI1Sq8

Length: 13min 14sec (794 seconds)

Published: Sat Dec 23 2023