How to become an All-Knowing AI and ML expert?

AI Discussions
1

Need urgent help and guidance

Dear senior AI experts,

I hope this message finds you in good health and high spirits.

I am writing to express my deep interest in the field of Artificial Intelligence (AI) and Machine Learning (ML), and my desire to become an expert in this field.

As a certified Project Management Professional (PMP) and Scrum Master from Afghanistan, I have successfully managed and monitored projects worth up to 300 million USD in my career. However, I am now looking for a career change that allows me to pursue my passion for AI and ML.

I have already acquired intermediate skills in Python and advanced knowledge in mathematics, which I believe will be an asset in my journey to become an expert in AI. Additionally, I have completed the “AI for Everyone” course and found it quite informative, which has only fueled my passion further.

As you know, there are numerous online courses available on platforms such as Coursera and Udemy. However, what I am seeking is guidance on the exact roadmap to follow to become an all-knowing expert in AI and ML (exact link to online courses and materials). I have watched many YouTube videos and reviewed various roadmaps, but I find that they tend to be general and lack specific details.

Therefore, I would greatly appreciate your guidance and suggestions on the specific courses that I need to follow in order to realize my passion for AI and ML. I am committed to pursuing this field and becoming a valuable contributor to the AI and ML community.

I am grateful for your time and look forward to hearing from you soon.

Thank you very much.

Warm regards,

6 Likes
2

Hey @Salah_hamidi,
Welcome to the community.

Let’s start with a small discussion about the above statement. I have been completely invested in AI for the past 2 years, and still, till date, I haven’t come across any person who is an all-knowing expert in AI and ML. I am not sure what “PMP” and “Scrum” exactly refers to, but I guess, it would be true in those disciplines as well. But if not, then allow me to tell you that AI is a highly dynamic field, and at present, even more so, and thus, it’s not possible for any single person to know everything about AI and ML. That being said, let me mention some great threads which you can find in this community, here for your reference.

In this thread, check out C.2 and C.3. There, you can find some threads, using which, you can devise your own road-map.

Now, in your thread, I can find an expression, “exact road-map”. Unfortunately, there is no such thing. The road-maps need to be developed by the learners themselves, since every person learns in a different way. Some learn via reading books, some learn via courses, some learn via blogs, some learn via Youtube, etc. So, any of us can only suggest the resources, the road-map is for you to create for yourself using those resources.

My personal recommendation for you would be to check out the Machine Learning Specialization, offered by DeepLearning.AI. I see it as the perfect hot-spot for beginners in AI, since it exposes a learner to the various sub-fields in AI. Once you have completed the MLS, it would be much easier for you to figure out the next steps in your AI journey.

Also, allow me to add some other threads, posted by learners along these lines:

Lastly, allow me to tag some other mentors as well, who have professional experience in AI, and perhaps, would like to add something to this. Hey @paulinpaloalto Sir, @reinoudbosch, @rmwkwok, @arosacastillo and @Juan_Olano, can you please share your advice to @Salah_hamidi. Thanks in advance.

Cheers,
Elemento

2 Likes
3

Hi Salah_hamidi,

The nice aspect of the Deep Learning Specialization you find here is that it takes you from the basic level of the AI for Everyone course straight up to the transformer architecture that forms the basis for current state-of-the-art large language models such as GPT, LLaMA, PaLM, BLOOM, Chinchilla, etc. Once you understand the transformer architecture, which is discussed in the final week of the final course of the specialization, you will be able to understand the most important aspect of such large language models as well as state-of-the-art object detection models. Your Python skills will improve steadily along the way due to the various practice assignments. If you want to be a bit more prepared before you start with the Deep Learning Specialization you can first take the Machine Learning Specialization.

Should LLMs be your main interest, then you can pursue this interest subsequently by taking the Natural Language Processing course. If you have a different focus and a practical orientation, the various TensorFlow courses will be useful to show you how to quickly build and apply models using TensorFlow. The MLOps course can teach you how to efficiently and effectively deploy such models using Google Cloud. If you prefer PyTorch and AWS and/or you have an interest in generative models, you can take the Practical Data Science specialization.

In other words, there’s much you can learn here - as I have done myself. In my case, the next step has been to dive into current literature on machine learning and transformers, whether on websites with papers such as arXiv, or in books by leading authors in the field. I have also started building a system I wanted to build for a long time, and am thinking about a start-up or cooperative venture.

As Elemento indicated, and as in any high tech field, you will never become an all-knowing expert. Things will always move very fast and you will always feel you are lagging in one way or another. Everyone has this. In my case, conceptualizing the system I am currently building has been a way not to be distracted by that. In fact, I experience it as highly stimulating that things keep moving fast and that there’s always something new to be excited about.

So, my suggestion is, start either with the Machine Learning Specialization or the Deep Learning Specialization, then find out what field of AI interests you most, and choose a next specialization that fits. Most importantly, enjoy!

4 Likes
4

@Salah_hamidi - I also had a similar question and still have it. You said you have acquired advanced knowledge in mathematics. How did you go about doing that? Any courses that you can recommend?

Also to others that are tagged here, is advanced knowledge in math is needed to be a decent level of ML/AI expert. Im asking because I started with the Machine learning speciality certification course in Udemy and not understanding concepts through that I am now enrolled into 3-4 courses just trying to understand the heads and tails out of ML concepts.

5

Knowing more math never hurts, but the detailed answer depends a lot on what your goals are. If you want to be a researcher at the leading edge of the field and advance the state of the art, then you probably do need to spend some time seriously studying math. You’ll need the equivalent of undergraduate level courses that a math major would take in linear algebra, calculus, multivariate calculus and matrix calculus at least. You should also take probability and statistics (the kind that math and physics majors take, not the kind of statistics that the psych majors take). There are online resources for most of these, but I have not really explored them, since I was a math major at an actual in person university “back in the day” :nerd_face:. I know it seems kind of old fashioned now, but I actually sat in a classroom and watched professors write equations and theorems on actual chalk boards and took notes. Then I went home and read text books and wrote out my solutions to the homework problems with pencil and paper. I don’t know whether such things still exist. :sweat_smile: One online course that is the equivalent of an undergraduate math course in Linear Algebra is Gilbert Strang’s MIT Linear Algebra course. The Khan Academy also has a lot of valuable math curriculum online, although the only ones I’ve personally looked at were advanced high school level courses.

But back to the question about goals: you can design and build solutions to ML problems using existing techniques without needing that much math background. You just have to be an experienced programmer and to learn the state of the art techniques. As Reinoud and Elemento have already pointed out, there is a lot of excellent material here from Deep Learning dot AI that will give you a great start on that journey. Start with the Machine Learning Specialization followed by the Deep Learning Specialization. Those two specializations are specifically designed not to require an advanced math background in order to understand. You just need a solid understanding of basic Linear Algebra: vectors, matrices and algebraic operations on them. You don’t need to know what an eigenvalue is or to have taken all of the Gilbert Strang course. You don’t need to know any calculus at all. Of course ML is based on calculus, but Prof Ng just gives you the formulas and you have accept them as the truth: you don’t have to understand how to derive the formulas in order to write the code to apply them.

There is also a brand new specialization from Deep Learning dot AI called Math for Machine Learning (M4ML). Not all of the courses have been released yet. You could take that either before, in parallel with or after the MLS and DLS specializations.

5 Likes
6

Thanks @paulinpaloalto for an insightful answer. I definitely dont want to be a researcher just a really good ML engineer or an AI/ML Architect. I have already gone into a lot of rabbit holes trying to understand what been put up on the slides and have enrolled into many courses just trying to understand it (w/o seeing any real value in it, most of them try to just dive into code). The MLS course here is really good in teaching concepts but some of those equations are really difficult to understand and I have learnt to just take them for face value. DO you think the specializations that you mentioned here are enough for those job roles that I mentioned?

2 Likes
7

Taking MLS and DLS are definitely great and useful steps to take on your learning journey. Just those two by themselves are probably not enough to land you a job. I should make an important disclaimer here: I do not actually work in the field of AI, so I’m not qualified to advise on what it takes to actually get a real paid job in the field. I did spend a whole career doing software engineering in other fields (operating systems, embedded systems) and was a hiring manager for a good chunk of that time and one concrete thing I can say is that just taking courses (even at famous universities like MIT or Berkeley or Cornell or Stanford) will at most get you the interview. You still need to be able convince someone in an “in person” interview that you really do know what you are talking about. That will require some real “hands on” experience solving problems with ML techniques. The courses will get you started, but then you need to seek out opportunities to apply what you have learned. E.g. by trying things like Kaggle challenges or finding other problems that are personally of interest to you and which you can figure out an interesting way to solve using ML techniques.

3 Likes
8

Thanks @paulinpaloalto for all the great answers really appreciate it. Sorry @Salah_hamidi for hijacking your original post, wasn’t my intention to do so.

1 Like
9

Hi everyone,

There is little to add here that was not mentioned or covered by my amazing colleagues. I will point out that, apart from creating an unnecessary pressure on your shoulders, you do not need to become a ML “god” to find a job :slight_smile:. Most companies they just look for people capable of learning, and creating+deploying prototypes. There are so many interesting ML areas to explore but right now the most promising one is on NLP so I would just to start with NLP courses and maybe contributing to some open source projects, such as in HuggingFace. This is already a big milestone, once you achieve that you can go on with other areas :slight_smile:

Just my two cents here.
All the best,

Rosa

3 Likes
10

You may find this helpful. It is a fairly decent comprehensive overview of specific skills required for Data Science/ML/Data Engineering/Deep Learning. It doesn’t link any course content but I found it useful to find areas to self study.

1 Like
11

@graham_broughton - Great reference, a bit overwhelming but awesome nonetheless.

12

Thanks! That was my first impression too :sweat_smile:

1 Like
13

Well mentioned, this field is dynamic and still evolving, there is always something new popping up here and there, another research paper, a new approach etc. so keeping an open mind all leveraging many sources is great