Should You Learn Machine Learning?

Welcome back to another video! Recently there has been a lot of hype around machine learning and AI and it seems like everyone wants to learn machine learning. So in this video, I will be discussing some of the pros and cons of machine learning and if you should learn machine learning.

📄 Resources 📄

⭐️ Timestamps ⭐️
00:00 | Introduction
01:52 | Machine Learning in Real Life (Story)
05:36 | Advantages of Machine Learning
09:53 | Disadvantages of Machine Learning

◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️
💰 Courses & Merch 💰

🔗 Social Medias 🔗

🎬 My YouTube Gear 🎬

💸 Donations 💸
◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️◼️

⭐️ Tags ⭐️
- Tech With Tim
- Machine Learning
- Pros
- Cons
- Learn Machine Learning
- What is Machine Learning

⭐️ Hashtags ⭐️
#TechWithTim #MachineLearning

[TheAwesomeness]

I would also like to point out that actually *understanding* machine learning well enough to tweak models effectively or read white papers on topics (as was mentioned as a benefit, being part of something), will be EXTREMELY difficult to do without a fairly high level of knowledge in linear algebra and probability + statistics. In fact, at my university, in order to enrol in the first machine learning course, you need to have taken 4 semesters of calculus (with proofs, as it highlights mathematical maturity one develops), one or two courses in linear algebra, and 2 courses in statistics.
That’s not to say it’s impossible to use ML without that, of course it isn’t. However, it is difficult to engage academically with the material without sufficient math knowledge (much more so than with required programming knowledge for ML).
So yes, you will be “a part of something”, in that as the field develops you might have fun new models to play around with, but it’s a bit disingenuous to suggest people will be able to comprehend ML publications at a meaningful level, or be generating their own custom models

[amadios]

Pros : It is fun

Cons : You have to work

Don’t thank me

[BGrovesyy]

I'm a data scientist and 95% of my job is data collection and preparation, it's still a cool job but it's not all model development

[AngelTaylorgang]

I read a similar story from the book “The Power of Habit”

I think it was Target

[chikanma]

the company tim talked about in this video is actually Target I learnt from Zach Star

[parmeet]

This is exactly what I was searching 😂 & I got this notification

[ananthramvijayaraj]

In my eyes, the second con is the worst thing that can happen to you in ml. In many cases I would eagerly wait for about 5-6 hours for the model to train, to see it not work at all in the way I intended. I'd keep training the model day after day after making small changes in the code to see its performance not getting better at all.

[UknownCompiler]

at first I thought i will learn machine learning in 6 months, with some coding experience. i dedicated 5 hours on normal day and 10 hours a day on weekends to study the field and make it in 6 months.
and guess what, in reality it took me 1+ year of hard working and studying this field to be able to build deploy deep learning models and to only begin comprehending state of the art models and begin to mix different algorithms to solve a specific problem.
time may vary from person to other but real hard work and dedication is required.
if i was to go back and change the way i learned things rather than just jumping to the field :
first to learn data structures and algorithms, programming logic.
then try to learn and apply some basic algorithms like binary search, reverse linked list, string manipulation..., this will help you so much to improve coding skills.
the first step will help you so much in processing data solving problems related to data and getting your models to work using your own way.
second learn some basic calculus and statistics, linear algebra, probability, dot product, gradient descent...
third learn machine learning algorithms and try to apply them using your programming skills.

[tntsharma]

Tim you should post more about machine Learning. Also please try to make a video on how to use the model for production. It will be really helpful!

[sudhamajayanthi]

Wtf, I thought ML Expert is a joke for a second until you showed the website 😂 Clemet is legend 🚀

[Limestarz]

I'm taking a few online ML tutorials and have a text book (not related to the course) and I have to say that it is very similar to what I did in econometric with regards to predicting. The regression analysis in ML is the same as I did in econometric but I just I didn't use python to do it, I was using Stata, but you could use R to do the same thing and now days I'm sure they are teaching python in economics. At any rate, even after taking econometrics and taking a ML course I still find it hard to use this knowledge for anything specific or maybe more aptly how to apply this knowledge to anything useful. Albeit I do recognize that ML is a large topic and the first part of learning ML that I am doing (including the textbook) all start off doing Regression Analysis and maybe that is what everyone doesn't use when they talk about ML and the other parts I have yet to get to like AI and Neural Networks is all the rage that everyone is talking about? IDK.

I often hear people say that ML is used for large datasets like at Netflix, Google, Amazon or etc to help suggest something (or something along those lines). Well, that's all fine and good, but I don't work at a company like Netflix or Google in terms of having that much data or that level of skill. The other problem is that to use ML at such companies is not really covered as they simple will say in the textbook or ML course that it is very complicated and won't go into. Thus, kind of making the example of these FANG companies almost useless in the short term, and medium term. When you learn about ML you often will be taught forecasting and you will be able to display it on a nice chart/graph with a line through it e.g. the trend line. Technically, you could use EXCEL to reproduce some of the things I have learned in ML, so at times feels kind of pointless and demotivating. Or, other things in ML like predict a single element if you have data on peoples age, salary, and whether or not they bought a car. You can then ask your data if someone at 30 years old and that makes $70k will they buy a car? It will be based off a percentage, of course but it will come out as a 1 or 0 indicating they bought it or not. The problem is normally at the end of all these tutorials the instructor will say something like the data you see in real life will never look like this and say this is for educational purposes. And real life application is much harder as it has more variables etc. So, it becomes hard to do your own projects in ML because they don't really explain what variables you should be including or looking at for your own projects.

Most of the time the ML tutorials focuses on how to do the predication/trend line and not care how you got the data and what data items you should be using. They just give you datasets from a website or make their own dataset. This is a huge issue if you want to try doing ML on your own data because they don't teach you what data you should be capturing or what practical purposes you could use this for on a smaller scale such as a small-medium business or say your website has x amount of traffic, etc.

I'm often left wondering how the ML regression Analysis chart that has a slope and line of best fit will be used in anything that they mention such as the things they say at the start of the chapter. For example, in my book they talk about using ML to create a spam filter and then you start off at the basics of learning about linear regression. Or they talk about using housing price data to predict how expensive a house will be based off like 5-7 factors. But to be honest the housing sample is a great example, but I often have trouble trying to think of a problem that I could create using ML to predict a result (ie my own project). Also, the examples often are intuitive, like yeah the bigger a house the more expensive its going to be I don't even need data or run ML on it to know that. It can get de-motiving at times. So, the examples are illustrator of basically what you already know to be true. Like predicting gas prices will be higher in the Summer time because more people drive because of more people take vacation/travel during summer. Also, they never mention how to implement that line graph into an application like the mentioned spam filter. It's 200+ pages like that. Countless hours of videos from tutorials saying, "Now, of course the real life data will not look anything like this, but you get the idea"... I am kind of stuck in-between smashing my head against the wall listening to the instructor on how a graph with a slope is ML and hoping that the next item they teach will get me to what I want. I'm tempted to skip 20 hours or 200+ pages ahead and dive into the Neural Networks but can't stop thinking that maybe Regression Analysis is going to be required or some of the tricks you are suppose to learn along the way will be helpful in the AI section, or whatever.

So far, it's like learning basic Statistics but in Python.

[cyberJali]

Thank you Tim . Really eye opener for me

[mohiths]

The only thing holding me back is that it took about 30 mins to implement a classification algorithm with a data size of around 20 numbers on a powerful laptop. It was straightforward and a lot of fun but when the size of the data increases, it might go on for even a week with GPUs. It pays well but so does frontend development at any decent company. This video didn't give any valuable inputs to make a decision

[ThePhantomCoder]

I LITERALLY just bought a udemy course on data science and machine learning... This guy knows our search history...

[joaosilveira]

I honestly don't get it why so many people hate math, it really is so important and fundamental to aalmoust anything, specially in programing and AI. I wich there were some more videos about math and programing

[ahmadhammad]

I knew it, you are using ML, that’s why most of your latest uploads “is reading our mind”.

[lucasrodrigues]

I have started learning flask and also thought I should learn machine learn and apply it to flask.

[kitgary]

Of course. ML is so hot and almost every companies expect their engineers to know at least some ML.

[emftechEE]

First time I’m this early! And excited for the topic!

[kenet]

Tim always posts videos on topics that I am currently practicing/have in mind. These coincidences happens so much, its scary lol.

Love your vids 👍