genAI vs ChatGPT vs LLMs - Buzzwords Explained!

Key Takeaways

greetings fellow Learners this is going to be the first video in a playlist of videos on gen AI but before we get into this wonderful world of gen AI I've got a thought-provoking question for you where do you see yourself using AI or gen AI in your life I like many of you use chat GPT in place of Google it's super fun getting into those philosophical rabbit holes with the not yet so sentient AI absolutely riveting now how do you use AI pray tell in the com ments down below and I would love to hear your thoughts now this is a great segue into today's wonderful video where we're going to talk about these buzzwords and explain their relationship with each other you don't need any prerequisite knowledge to understand the video and we'll explain the terms at a high level as we draw out a VIN diagram to show their relationship with each other this video will be divided into three passes with multiple quizzes along the way so pay attention get them right make your mom proud let's get to it let's start with our first buzzword artificial intelligence artificial intelligence is a field of study that involves developing methods and techniques to solve problems so in this case the field of study is AI artificial intelligence and one important term that we also want to describe here that'll come up throughout the video is that of a solution or in this case an AI solution an AI solution refers to a specific method or technology that is applied to solve a problem so what's an example of an AI solution let's say that we want to build a system that takes in a speech wave and determines what word you say this is a famous problem of speech recognition now one solution is to solve this problem using some rule-based system so we break the Soundwave into chunks map them to phones and recognize the word spoken now this solution is an algorithmic method to solve the problem and hence by definition of an AI solution this rule-based system is an artificial intelligence solution so right now let's represent the field of artificial intelligence with this large rectangle over here and within these fields we're going to write down individual Solutions so in this case it would be like a rule-based AI system the next buzz word is machine learning machine learning is a field of study that involves using a trained model to solve problems so the field of study here is machine learning and then the machine learning Solutions are trained models a trained model is a method an algorithm or technology that is developed by analyzing patterns in data the field of machine learning is a subset of the field of AI this is because the machine learning Solutions which are train models are a subset of AI Solutions too and because machine learning is a field we draw it as this rectangle so what are examples of solutions that fall in the outer rectangle versus this inner rectangle so let's take the rulebase system we described previously the solution used was a bunch of rules to recognize a word this is not a trained model based on patterns in data and hence rule-based systems are AI Solutions but not ml Solutions and hence they lie in this outer region but what's an example of a solution that is an ml solution well instead of a rule-based system we can train a model on data to recognize patterns between speech and different phones and so we could take a speech wave process it and then convert it into phones in order to predict a word and the model here could be a hidden marov Model A support Vector machine and many more because these Solutions were created from analyzing patterns between speech and words the solutions are ml Solutions hence they fit in the inner region the next buzzword is deep learning so deep learning is a field of study that involves using trained neural network models to solve problems now neural networks are interconnection of neurons and the field of study here is deep learning the solution is the trained neural network models and these are a subset of train models and so deep learning Solutions will be a subset of ml Solutions and so this deep learning field will be a subset of the ml field and hence Because deep learning is a field of study we can represent it as a rectangle once again and that rectangle is going to be embedded within the ml rectangle now let's write the solutions again so the rule-based systems are just AI not ML and hence they're also not deep learning hidden Markov models support Vector machines are trained models but again these models are not neural networks hence these Solutions are in the ml region but not in the Deep learning region now for a solution to fall in this deep learning region we would need a neural network that is trained to analyze patterns between speech and words and then we could use the neural network to take in some speech and predict the spoken word quiz time have you been paying attention let's quiz you to find out which of these is a machine learning solution a a neural network to classify an image as a cat or a dog b a fixed formula to determine the price of a house given the last sale price c a decision tree created from data to classify an image as a cat or a dog or d a decision tree created based on intuition to classify an image as a cat or a dog note here that multiple options may be correct and I'll give you a few seconds to answer this question the correct answers are A and C but can you tell me why give your reasoning down in the comments below and I would love to hear your thoughts and at this point if you think I do deserve it please do consider giving this video a like because it will help me out a lot that's going to do it for quiz one and pass one of this explanation but keep paying attention because I'm going to be back to quiz you [Music] now let's talk about natural language processing or NLP NLP is a field that involves using methods and techniques to solve natural language problems so the field here is natural language processing and the solution can be any method or technique so by this definition NLP Solutions are a subset of just AI Solutions but that's where the comparison stops because NLP Solutions may or may not be trained models hence the rectangle for NLP could intersect with ML NLP Solutions may or may not be trained neural network models hence the NLP rectangle can intersect with this deep learning rectangle 2 and so because NLP is a field of study we can draw the rectangle accordingly so what are examples of NLP Solutions now honestly it could be the same as all the AI Solutions as long as they solve language problems so A rule-based system that solves speech recognition falls under NLP But A rulebase system that predicts the price of a house falls under AI but not NLP hidden markup model support Vector machines are designed to solve speech recognition that we described before and these fall under both NLP and machine learning support Vector machines designed to class ify images let's say as a cat or a dog can fall under machine learning but not NLP neural networks that solve speech recognition fall under NLP and deep learning but neural networks that predict the price of a house fall under deep learning but not NLP I hope these bucketization of solutions are still looking pretty clear now the next term that we want to do is the Big Kahuna which is Gen a I so gen is a field of study that involves using generative models to solve problems so the field is geni and the solutions are going to be generative models now we discussed before that a train model is a method algorithm or technology that is developed by analyzing patterns in data now a generative model is a train model that learns the distrib distribution of data and can generate new data so by this definition gen AI Solutions are a subset of machine learning Solutions this is because ml Solutions are trained ml models gen Solutions are trained generative models now gen Solutions may or may not be trained neural network models and hence they intersect with this deep learning rectangle now because gen is a field we represent this with a region on the V diagram so what are examples of gen AI Solutions gen AI solutions that is just Ai and ml Solutions and not deep learning Solutions include marov chains so to paint this picture let's say that we want to create a system that completes a story based on the given first 10 words Marco chains are models train on the text databases to understand how to generate text and hence they are trained generative models and hence marov chains are gen AI Solutions but they are not neural networks so they are not deep learning Solutions also this specific marov chain because it solves language problems also falls under an LP but we can also have markup chains that are trained to generate music and these would fall outside of the NLP rectangle but still in gen aai just to note here that markup chains they're not really used that much these days with the Advent of larger more powerful technologies that we'll discuss next but just to understand for the sake of completeness you can see how they fit in quiz time it's that time of video again have you been paying attention let's quiz you to find out which of the following is an example of gen AI a a neural network that can generate images b a neural network that can generate text c a neural network that can classify an image as a cat or dog or d a decision tree that is Created from data to classify an image as a cat or a dog note that multiple options here may be correct I'll give you a few seconds to to answer this question the correct options are A and B but can you tell me why comment your reasoning down below and let's have a discussion that's going to do it for quiz time and pass two for now but keep paying attention because I'm still going to be back to quiz you next is language model so a language model is a method or technology used to solve a language modeling problem so here the field is broadly natural language processing as language modeling problems are a subset of language problems and the solution here is going to be some method or technology that is the language model so a language model takes in context words and it predicts the next word or some middle word now from the definition of the field the language model is an AI solution it is also an NLP solution as the language model solves a very specific language problem and so these Solutions can fall in either machine learning deep learning or just generative AI so for example the the mark of chain that we discussed in the previous pass is a language model because it takes in 10 context words and it tries to generate the next word now we can come up with a rule-based language model that predicts the next word based on rules and this would be just AI because there's no learning of patterns in data now a language model that makes use of a neural network is a deep learning solution and an example of this would be large language models so introducing large language models is a buzzword the large language model is a large large neural network used to solve the language modeling problem so large neural network is a neural network with just a lot of parameters to learn typically in the millions or billions of parameters the field is natural language processing and AI as large language models are language models and they also belong to the same Fields now the solution here is a large neural network and hence llms large language models are a subset of deep learning Solutions hence we can say that large language models are a class of Solutions in the Deep learning and NLP intersection what are some example Solutions let's just say of these large language models an example of large language model that intersects with Gen is Chach PT so chat GPT is a neural network with a large number of parameters and hence it is a llm chat CPT is trained to understand the distribution of text Data from the internet and in usage we give it a question it generates text and hence chat GPT is a generative model and is a gen AI solution too now what are the examples of large language models that are not gen well this could be a Bert model that is used as a sentiment analyzer for example we could pre-train Bert on the Tas of language modeling and then we could fine-tune Bert to classify an input sentence as being a positive sentiment or a negative sentiment and we could do the same for GPT as well and so this here is a more complete diagram of the fields as rectangles with their Solutions laid out inside quiz time ooh this is going to be a fun one which of the following is an example of gen but not a large language Model A a large neural network chatbot that generates text based on a question b a small neural network that generates summaries of input documents c a logistic aggression that classifies a movie review as positive or negative review or d a large neural network that classifies a movie review as positive or negative review note here that multiple options may be correct and I'll give you a few seconds to answer this question the correct option is B but can you tell me why comment your reasoning down below and let's have a discussion and at this point if you do think I deserve it please do consider giving this video a like because it will help me out a lot now that's going to do for quiz time and pass three of this video but before we go let's generate a [Music] summary in this video we explain some important buzzwords in the AI space we identified the rectangular regions as the fields with some specific solutions that we have written in them and with these definitions we've created this beautiful diagram and so I hope this adds Clarity to how these different terms are related to each other and that's all we have for today if you're interested now in learning a little bit more about some technical details on say reinforcement learning which is a core learning Paradigm of machine learning I have a playlist of videos that you can check out right over here thank you all so much for watching if you think I do deserve it please do consider giving this video a like and I will see you in the next one bye-bye