AI From Scratch, Part 1: Start Here Your Journey In AI

AI From Scratch, Part 1: Start Here Your Journey In AI

Back when I started learning about AI, things were “different”.

There were a few trustworthy resources everybody was talking about. It was easier, I think, to know where to go if you wanted to get started learning about AI.

There was a guy in particular, prof. Andrew Ng, that kept popping up in forums, under YouTube comments. He was everywhere and for good reasons.

But I didn’t know who he was back then.

I only knew everybody was suggesting his Machine Learning course on YouTube, a course often referred to simply by its code at Stanford University: CS229.

On the other hand, the field of AI had a really long way to go. There were tons of good ideas, but they all were at their infancy.

Teaching AI has also changed a lot, like everything else during these years, and that’s what I want to address in the course of this series.

I’m going to share with you the best material I know on how to learn about AI, the ones I would use if I were to start all over again.

I will also share some general learning principles I’ve adopted during the years that made a huge difference in getting the most out of what I study.

I’ll invite you to answer questions and explain concepts in your own words, to test your understanding.
I will also suggest resources that include intuitive understanding of concepts, in-depth theory and practical hands-on in order to get a complete learning path.

I assume you already have your reasons to learn about AI and I’ll move directly to how you can start learning it in practice.

Finally, expertise in AI takes time, effort and continuous learning.
I’ll strive to give you all the necessary knowledge and resources to effectively embark on this journey, but expertise in AI requires dedication.

Don’t feel the need to rush. It’s not going to help on the long-run anyway.
With that in mind, let’s begin our introduction to AI.

“AI For Everyone” By Andrew Ng

As I said, a lot has changed in how AI is taught, and there are many more resources available to you now.

CS229 was straight into the maths, in the context of a University degree in Computer Science.
Now we have more gentle introductions to AI, but the path I’m going to suggest you, starts from the same professor.

“AI For Everyone” is an introductory course explicitly targeted for non-technical people.

This works equally well for:

• Beginners that plan to get deeper in the technical aspects of AI.
• Professionals that are considering to introduce AI in their companies (e.g. managers, CEOs and other decision-makers)

It is hosted on an online platform called Coursera and taught, of course, by Andrew Ng.

But who is he?
Prof. Andrew Ng is one of the most well-known AI experts and educators in the field.

He’s a professor at Stanford, researcher and entrepreneur.
He is the founder of deeplearning.ai, Landing AI and co-founder of Coursera itself, the platform where the course is hosted. 6

I can go on: He was chief scientist at Baidu, co-founder of Google Brain… I’m sure you’ve got the point.

Coursera was co-founded with Daphne Koller – also professor at Stanford – with the vision of democratizing AI by teaching it to as many people as possible. 6

Over time, this vision took Andrew and his team to develop additional courses aimed at non-technical audiences interested in AI.

That’s where “AI For Everyone” comes from.

By the way, the course is completely free to audit.
This means you’ll have free access to all the videos, but if you’re interested in a certificate upon completion, you are required to buy the course. 2

To give you a sense of what topics you can expect to learn about, here’s the four week syllabus of the content:

• Week 1: What is AI. You will learn the meaning behind common AI terminology, including neural networks, machine learning, deep learning and data science. You will also understand what AI realistically can and cannot do, and what makes an AI company.
• Week 2: Building AI Projects. This covers the workflow of a machine learning project and a data science project. You will also better understand how to choose AI projects to tackle, how to work with an AI team and what technical tools are often needed to them.
• Week 3: Building AI in Your Company. You’ll learn how to build an AI strategy for a company. You will also see some case studies of AI applications, such as smart speakers and self-driving cars. You will also get to know common AI pitfalls and how to avoid them.
• Week 4: AI and Society. Here you’ll see how AI is impacting society and what you can do to navigate ethical and societal discussions surrounding AI. You will also learn about some of the challenges and risks of AI, such as discrimination, bias, adversarial attacks and adverse uses. You will also learn how AI affects developing nations and jobs.

I’ve followed this course myself, out of curiosity and to understand whether it could be a good fit for a beginner in AI.

Sure I think it is, but I want to highlight this is an introductory course to AI.
You should expect to come out of it with a good sense of what AI is, but not to get any technical understanding out of it. That’s not the purpose of this course.

We’ll start talking about the technical aspects of AI, in the next post about Machine Learning.

I’d also say some of the examples provided about AI capabilities, already feel ancient, as “AI For Everyone” was first released back in 2018.

While the core content remains as relevant as it can be, the examples suggest how rapidly the field is advancing… which is good!
On the other hand, it’s easy to feel “left behind”.

That sensation of “missing out”, can be easily used to sell you doubtful “free ebooks” and AI tools, promoting a false sense of urgency. 7 8

My advice is to completely disregard those feelings, which are normal but often used against you, and focus instead on building foundational knowledge over time.

The rapid evolution of AI should feel as a chance to be part of something significant, not a reason to worry about keeping up with every change.

Shape Your Environment For Success

“AI For Everyone” is surely a wonderful resource, but to truly learn something effectively, you want to fully immerse yourself in it.

Shaping your environment, means paying attention to what you are surrounded by most of the time, and make proactive changes so that it is easier to achieve your goals.

In the context of learning AI, this has a number of positive effects 3 4 5:

• It makes your brain more receptive to the concepts you will be exposed to. When you will study them in a course, you’ll be ready to learn them.
• You’ll notice which topics come up more often. That allows you to prioritize important ones and understand which are foundational in AI.
• Keeps your motivation high as you’ll see the amazing things that happen in the field and what is being built around AI research and tools.

With “environment” here I mean everything you’re regularly exposed to and that influences your behavior. 1
That includes the people you spend your time with, the experiences you have and, generally speaking, all kinds of different “inputs” your brain gets on a regular basis.

I can’t tell you which people you should spend your time with, but you probably use a combination of social media and emails on a daily basis.

That’s our target.

Of course, if you can also spend your time with AI researchers, undergrads or even just AI enthusiasts, that’s part of the idea as well.

Meanwhile, the newsletters and YouTube channels below, are a collection of high-quality resources you can use to regularly expose yourself to AI.

AI Newsletters

I’ve personally started reading AI newsletters quite recently.
After ChatGPT and AI Image Generation, the amount of AI related news was overwhelming and I genuinely had difficulties at keeping up.

The two newsletters I suggest here – “The Batch” and “TL;DR AI” – have solved the problem for me and are a way to keep up to date with trends, research jargon and important AI news.

I personally find they well complement each other.

In particular, I would argue The Batch focuses on depth, commenting over a few important news but going deep into each analysis, while TL;DR AI focuses on a broader coverage of AI news and announcements, making it easy to remain up-to-date.

The Batch, by Andrew Ng and deeplearning.ai

“What Matters in AI Right Now” – the tagline of The Batch – may be an overstatement but the most important news are really discussed in-depth.

I found The Batch while exploring the deeplearning.ai website in search of courses from Andrew Ng (I know I seem obsessed, but I love how he teaches) and found this newsletter.

I couldn’t refrain myself to subscribe.

Each release of The Batch begins with a letter from Andrew Ng himself, followed by a coverage of the main AI news of the week.

Every news is summarized in its main highlights and commented thoroughly on why it is relevant for the field and what impact it can have.

The fact that it is weekly, means only the very top news of the week are covered and other announcements are not part of the scope of the newsletter.

There is so much going on in AI these days that it’s hard to cover everything that “matters”.
That’s why I also want you to suggest TL;DR AI.

TL;DR AI

Different things will matter to different people and TL;DR has a broad coverage of topics.

It is a daily digest of major AI news, research articles, tools and a variety of other resources.

Each of those is linked and briefly summarized to give you context and let you decide what merits your full attention.

This newsletter, by itself, accounts for much of what keeps me up-to-date with current AI news and trends.

AI YouTube Channels

In our quest for finding the best introductory material to AI, visual resources help provide a complete picture.

The purpose of these channels is to show you AI from different angles and add variety in what you are being exposed to.

Among these channels there is research-level material, AI news and intuitive explanations of AI algorithms.

The people behind them, are anything from AI researchers to chip design engineers, which I think provide a healthy diversity of different perspectives.

Yannick Kilcher

I’ve known Yannick Kilcher’s channel for as far as I can remember.

He is an AI researcher that breakdown AI papers and I know him mostly because following the jargon and maths of AI research papers is often a mess, and he makes them much easier to follow.

Yannick’s “ML News” series also provides a commentary upon recent big announcements, projects and other news in AI.

I’m not sure he follows any regular schedule for ML News, but when it comes out it’s always an interesting perspective.

Anastasi In Tech

Anastasi is a chip design engineer and in her channel, she mainly covers AI breakthroughs related to hardware improvements.

Hardware is such a big component of AI but when people think of AI they mostly think of neural networks, chatbots or whatever new model/algorithm makes the news.

All of that has to run on hardware and a big reason modern AI exists, is because there now is the hardware needed to support all those computing requirements.

I follow her channel to keep an eye on an aspect of AI that is fundamental, but I tend to not read much about.

StatQuest with Josh Starmer

StatQuest is one other channel, along with Yannick’s, that I dearly followed during my University degree, and for the same reason.

Having someone more knowledgable than you, that is also able to teach things so intuitively, is a godsend.

The focus of StatQuest are AI algorithms of all sorts, from Machine Learning

Everything on StatQuest has the purpose of making you approach the theoretical aspects of AI in a way that isn’t intimidating (at all).

After you’ve seen your first video, you’ll understand the guy has a gift at explaining hard concepts in very simple words.

I often rely on these videos when I want to learn or refresh some algorithm because his style just gets into your head and makes concepts so much easier to understand.

He also started a podcast – “The AI Buzz” – which is released about twice a month.
This can become part of your regular AI news to keep yourself up-to-date along with TL;DR AI and The Batch.

AI Explained

Philip’s “AI Explained”, is one of the many AI channels that emerged after the arrival of ChatGPT.
Most of those channels are improvised AI experts that jumped on the train when they saw the occasion.

This channel is on a different level.

Philip comments AI news and research through great analytical skills and clarity in explaining his conclusions.

I know many other recent AI channels, but AI Explained is on this list because I sincerely find him a great and lucid thinker at par with the others in this list.

Thus far, I have suggested you a structured course like “AI For Everyone” as well as a collection of less structured but in-depth resources that will provide you with a variety of different perspectives.

In the next section, I will suggest you how to put all these resources together and what to pay attention to, in order to get the most out of it.

How To Best Use All The Material Covered So Far To Learn AI

Some quick steps you may want to do quite soon are:

• Create an account on Coursera and enroll in “AI For Everyone“.
• Subscribe to both TL;DR AI and The Batch
• Subscribe to the YouTube channels suggested before: Yannick Kilcher, Anastasi In Tech, StatQuest, AI Explained

AI For Everyone + Life-Long Learning

To get the most from “AI For Everyone”, consider doing the following:

• Take (very short) notes of the highlights of each week. Don’t waste your time into lengthy notes, but don’t even overlook taking them.
• Take your time to digest and think through the new concepts that are being presented to you.
• Occasionally stop and try to explain in your own words the topics you’ve seen so far.

I’ll cover learning principles in the future, if you so like, because life-long learning skills are something valuable in a fast-changing field like AI.

For now please remember that, although challenging and even awkward at times, explaining topics in your own words and asking yourself questions are among the best ways to build a deep understanding. 5

Start getting into the habit of doing these things every time you want to learn something new.

An example of a relevant list of questions for AI For Everyone may be:

• What is AI? How does it compare to Machine Learning and Deep Learning?
• What is the difference between Machine Learning and Data Science according to the course?
• What should you be paying attention to in order to ethically develop AI systems?
• What are reasonable use-cases of AI and which are not?
• What is an “adversarial attack” and how can AI systems be challenged by them?
• What is a common workflow for Data Science and Machine Learning projects respectively?

Get Comfortable With Research Papers

Even if you are just starting out, start engaging with research papers pretty soon.

Reading only the Abstract and Conclusions is enough, but don’t get intimidated by research.

Some research papers are so full of jargon that even full-time AI researchers spend quite a lot of time understanding them (Yannick may help in those cases).

Other papers instead, are much easier to follow and packed with information.

The world of AI research is a huge one and there’s plenty to learn from, regardless from whether you want to become an AI researcher or take any other career path.

Yannick’s channel along with the papers included in TL;DR AI and The Batch are a great place to start.

Don’t Limit Your Learning to Formal Courses

Here I covered AI For Everyone as a first introductory “structured” course, but you may as well be a University student in a Computer Science or AI degree.

Whatever the case, don’t stop there.

Explore topics that interest you by going through your own searches as well. Allow yourself to occasionally get lost in rabbit-holes out of curiosity.

Follow AI experts, YouTube channels and newsletters and learn from those as well.

As you keep learning about AI outside of structured courses, you will surely encounter unfamiliar terms that won’t be explained to you right away.

That’s alright.
When it happens, take a note to remind yourself about them and notice how frequent that concept is.
When you’ll later encounter a frequent concept during your courses, be sure to pay special attention.

If “Transformers” start to interest you before learning about gradient descent (a foundational learning algorithm), that’s fine!

You’ll eventually get to that, but cherish your curiosity. Let it motivate and guide you.

Exposing yourself to new concepts early, will prepare you for when they’ll be covered in depth later. 4
Don’t fixate over a pre-defined path, and get comfortable moving around by yourself.
This will make a huge difference in your career.

Final Words of Advice

I think I’m giving quite uncommon advice, but… take it slow!

AI is here to stay.
It’s a growing and healthy field, everything’s fine and you need not to rush to get it all and now.

Learning requires time for your brain to “rewire” and adapt to new knowledge. 5
Fast learning is also fast forgetting, rushing won’t help you.

Instead, take your time to practice, experiment and explore topics even just out of curiosity.

I say this to contrast a popular but unhelpful tendency to overwhelm you with the largest amount of AI courses, books and articles, without emphasizing how learning requires deliberate and focused practice on a manageable amount of material.

Rushing will leave you repeating the same material over and over again, with little real gains for your knowledge.

In the next post, we’ll look into Machine Learning through an approach similar to the one used in this article.
I’ll first introduce you some material to learn from and then suggest how to best use it.

It may not come as a surprise, we’re also going to discuss Maths and Programming as two main pillars on which AI is built on.

Go complete “AI For Everyone” and when you’re ready move to the next post:
[AI, Part 2: Make The Most Out Of Andrew Ng’s Machine Learning Specialization](TODO LINK) (Right-click and Add To Bookmarks)

If you enjoyed this article, be sure to [subscribe to The Harbor](TODO LINK) so we can remain in touch.

Extras & References

What do I mean with “Environment”

To dig deeper in what I refer to with “environment”, I find struthless’ video to be on point.
Watch the video and you’ll notice the connections with this article.

• 1 How a space can improve (or destroy) your life – struthless:

How does the environment of an AI learner look like?

• They probably are subscribed to many different AI related resources (maybe a few newsletters and some YouTube channel?)
• They study AI both from courses and from web searches and high-quality videos driven by their own curiosity to learn more even outside courses.
• They may have a few AI books around to read (we’ll have them in the next posts of this series).

That’s surely what I do myself!

Keep in mind what an AI learner – as you are now – would do.

Also, pay attention to friction: having newsletters and videos so easily presenting new with up-to-date AI news, is a way to remove friction.
Rather than you needing to do multiple searches to get in par with what is going on, this is neatly presented to you in an easy-to-read format right in your inbox or YouTube feed.

Should I Get A Certificate For “AI For Everyone”?

I have detailed advice about Coursera certificates here:

• 2 [Should You Get Coursera Certificates?](TODO link Coursera certificates post)
  In the article I make a point that some courses give you more value for your money than others.

In this case, paying gives you the certificate and access to graded quizzes for each week. There are no graded labs or extra content, though, which is instead the case for other courses.

Ask yourself: do you have any professional advantage in getting this certificate?
Maybe you already are a professional and this can give proof to your employer of your efforts in getting to AI related roles.

In that case, though, you may get your employer to pay the course for you. If they are not willing to, maybe they don’t even value the certificate that much in the first place.

But if you are a student – and maybe you’re not swimming in money – I don’t see how paying for this can be helpful now.
Don’t feel pressured to buy it, if you don’t have a clear reason.

You either know the certificate gives you a good advantage or you find someone that pays it for you (e.g. your employer)

Final disclaimer: I support my work here mainly through affiliation.
If (and only if) you already know you want to buy the course, and you decide to do so through a link here from the Harbor, you also happen to support me as I get a [commission back from Coursera](TODO link to specific post on my relationship with affiliate programs).

It’s a great way to support this project (with no extra cost for you, of course).

More On The Principles Of Successful Learning

In “Shaping Your Environment” I’ve mentioned how being exposed to the keywords of AI, “prepares” your brain for when you’ll learn those concepts later on, maybe on a structured course (like “AI For Everyone” maybe?)
I also mentioned the importance of motivation in learning.

These resources provide deeper explanations and evidence for those claims:

• 3 Motivation in Learning, Filgona et. al – Researchgate: Research paper. How motivation influences learning.
• 4 Three Ways To Prime Students For Learning – Faculty Focus: An overview on Priming and how to do it for learning.
• 5 Make It Stick: A book on the principles of successful learning. I’ve read this fully. It has shaped the way I think about learning and I use these principles daily. I also refer to them throughout this full series about learning AI.

Other References

• 6 Andrew Ng – About: Some background on Andrew Ng and Coursera.
• 7 FOMO Definition – Wikipedia: The general definition of what is the Fear Of Missing Out.
• 8 How To Stop AI FOMO – The Algorithmic Bridge: A specific in-depth discussion on FOMO in AI.