A Quick Introduction to Deep Learning by @misraturp

foreign this is mustra and today I'm going to give you a three minute introduction to the state of deep learning today so let's start by positioning deep learning in the general scheme of things so in the world of computer science where does it fall so deep learning is a specific approach or specific group of approaches and algorithms in the class of machine learning and machine learning is a specific approach to achieve artificial intelligence and artificial intelligence is a branch of computer science so that's how these things are related to each other kind of when you think about them in a bit like nestled way so why do we use deep learning and why do we see a lot of advances in the world of AI especially in the world in the way of deep learning lately so you know this year we've seen uh Dali 2 very recently chat GPT before that gpt3 or you know imagine all these image Generation video generation models or maybe text generation models why do we see them with deep learning so the main reason is compared to traditional machine learning deep learning has some advantages for example you can work with abstract problems you do not have to Define your problem as rigidly as you have to do with machine learning algorithms and also you can work with more unstructured data you do not have to create columns or tabular data from the data set that you have in front of you to be able to feed it to deep learning you can use unstructured data for example like images and lastly deep learning algorithms are able to understand the patterns inside this data just by directly looking at the data itself so that really lowers the amount of feature engineering that you have to do that normally we do with traditional machine learning but there are cons to training deep learning models too for example they take much more time they require much more computational power and many more times the data to be able to create an accurate model and the reason that we're seeing models like chat GPT or Dolly tubing produced by these big companies is simply because they can afford the computational cost that comes with training big models like this there are many different techniques or approaches or algorithms in the area of deep learning the first one was neural networks and most of the techniques that are being used today are mainly based on neural networks and then we had this era where everyone was using cnns and rnns and maybe a bit more of advanced versions of rnns for example like lstms but they're not really being used that much anymore just because there are so many other techniques and architectures that are out already that perform much better for example we have Transformers or diffusion models but what you have to remember is these really Advanced models which will probably also be outdated in a couple of years are using little components from their ancestors so when you look into a Transformer sometimes it is using something some logic that we learned back in the time from Iron ends or CNS or even all the way back from neural networks so that's why it makes sense if you want to learn about deep learning it makes sense to go back to neural networks and kind of understand the fundamentals of how they worked how they learned and how they were trained if you want to go further and understand how neural networks are trained how they learn and also get a more in-depth look at the fundamentals of deep learning you can take a look at my booklet called fundamentals of deep learning in 25 Pages through the link in the description I hope this video was helpful to give you an introduction into the world of deep learning thank you for tuning in and happy New Year foreign