4 reasons you should learn Machine Learning now

Most of you at some point of time on in last few years would have made an attempt to learn machine learning/ AI but may been discouraged by the sheer number of courses available on internet. A big number of those courses take the coding route without really explaining the underlying concepts. I must admit a number of "Foundational math courses for Machine learning" are equally off putting. Plus there are folks who may be looking for a conceptual understanding of AI and its application challenges without getting into the nuts and bolts of ML like those in leadership roles or domain experts but do not have much options. Until I completed the Machine Learning Specialization by Andrew Ng on Coursera .

After years of looking for a right introductory course, I was prepared to sweat it out, come what may. But I am pleasantly surprised how easy, well structured and well defined this course is. This has been designed keeping in mind the requirements of people who needs a well structured and conceptual introduction to AI but also meeting the needs of people whoc may be wanted to develop their careers in technology.

I thought of sharing my experience of doing this course as well as make a case for others to start learning Machine Learning.

What can you expect in this article?

?1.????Changes in job market driving need for better knowledge of AI and its application

2.????Why is now a good time to start pursuing Machine Learning Specialization

3.????The barriers to pursuing ML education and how to overcome them

4.????Resources for planning your ML education and keeping abreast on latest

?

??

I have been following AI and its potential impact on my career since 2014. At the time, there were either literature which were borderline scientific paper or videos of unreal AI fantasies at best. A lot of media was focused on reporting advancements made by big companies like Googles and Amazons on complex problems surrounding language models and early stages of computer vision. News of computers being able to recognize images of cute cats flooded people’s imagination. The only route available to learn AI was to first master Python coding and then learn to use various Python libraries. There wasn’t much of an easy start to AI then. ???

?

I am a Supply chain management consultant and my field involves forecasting and optimizing various operations of retailers, life sciences companies, manufacturers, etc. The global nature of operations means companies at any point of time are using 200+ commercial off the shelf software products to orchestrate these operations. The last generation of such products were in fact doing exactly the same stuff that machine learning models now are capable of achieving in a shorter time as well as more intelligently. The difference was the compute power available then was a fraction of what is available today. So, for me personally, as soon as I heard about AI, I wanted to learn and understand ML to be able to apply it to the business scenarios my customer face in their job day to day.

?

Not all industries’ and not all business and government functions’ data will be easy to generate or capture, especially in cases where human emotions are involved such as matters of legal judgement. The AI still has a long way to go for these functions to avoid unwanted impact of bias but anything in between the two extremes of industries will benefit greatly from application of AI.????

?

My curiosity forced me to take the coding path. I did take Python coding courses but the AI foundations were still missing. The books, YouTube videos and coding courses were just making it difficult because there wasn’t a coherent and well-structured course available that could take could combine all the math, coding and concepts together to make sense of AI. Those courses did not necessarily bring me any closer to unraveling AI from a layman’s perspective. If I remember correctly, in 2016 Andrew Ng, the Founder of Google Brain and Professor at Stanford university launched the Machine learning specialization course. That was one of the best courses and quickly became #1 course for professionals and student with non-coding background to really understand machine learning.

I recently completed the latest version of this specialization on Coursera. And it really got me motivated to write this article. I think the course is at a level now that provide the foundations and brings all the disciplines together to help learners really get a good grasp of the topic. I believe all must take advantage of this course and so thought of writing this article to remove any barriers you may have. I’ll cover 4 areas in this article:

?

1.????4 reasons why people should take this or similar course on Machine Learning

2.????Where does this course fit in our seemingly different career journeys

3.????Taking the lid off any mental barriers on this course

4.????And finally, some links of channels (Podcasts, videos, etc.) that describe AI with is real world application and challenges in plain English

?

?

?

04 Reasons you should do Machine Learning Specialization now

?

Let us address the most fundamental question - “WHY”.

?

1.????Nature of Jobs is changing:?The field of AI has the ability to transform the society the way wheels, electricity and industrial revolution did. This saying somehow seems old now but in reality, the effect of this change is already beginning to be felt. There is a widespread skill shortage of right digital skills even today and jobs related to AI continue to be paid highly.?

?

AI is impacting all types of businesses and facets of business in a transformative way. At this stage, the businesses which have traditionally been collecting and utilizing data in a structured format has seen the highest adoption of AI but techniques and infrastructure is increasingly being deployed to collect data where it was not collected traditionally. 5G will increase data collection capability 100-fold in the next decade. AI applications we see in the market today are examples of narrow AI – A small focused area application where AI is doing one or a small set of activities. Companies and researchers are continuously stretching the boundaries of AI and associated hardware to increase the scope and complexity of problems AI can solve. Way more AI talent is required to identify problems suitable for addressing with AI once the scale and availability of data increases in the near future. ???????

The jobs in the next 2 decades are going to look much different. We are increasingly seeing use of various digital technologies. Today the focus is primarily on Cloud. Companies are moving data from siloed systems and Databases to cloud and utilizing the cloud services and applications to drive changes en masse but much faster and at a fraction of cost compared to the traditional on-premise models. Once companies of all scales and sizes have shifted their data in cloud and put in place the right infrastructure to receive loads more data on it, the importance of AI will skyrocket. In most organizations today, due to the nature of IT landscape, data is split across hundreds of different software applications but once this data is available in one place in cloud, AI models like Deep learning or Deep Reinforcement learning could generate value for a company never seen before. What I am writing here is not even scratching of the surface of the change that is coming our way.

?

Understanding how these changes will unfold in your industry, learning what ML models and how best to deploy them will be crucial skill sets organizations of (near) future will require.?

?

There will be need of people who understand a business deeply (domain experts) but also speak the language of ML to help developers and tech community develop the right ML solutions.

?

?

2.????From Business led value to Technology led transformations

?

In my job I solve customer’s problems and I think until the beginning of last decade, value came from business strategy and IT was considered expenditure. Not anymore. Today, a company not in the forefront of digital technology doesn’t stand much of chance to compete in the marketplace. Even today, the growth discussions ride heavily on right digital innovation strategy. From being a tool of business enablement, Digital Tech has taken centerstage of enterprise and government value generation approaches. And AI is at the core of this approach. Cloud is an infrastructure technology; it is and it will pave the way for enterprises and governments to harness the data but value will be uncovered by AI. AI looks for patterns in data. Data collection is on rise but is still quite limited in scope and may not capture the patterns outside of a business function, ability or scope within which it operates. Imagine once data is collected by every device around us at every stage of a human interaction, the patterns and insights generated by AI will be immensely valuable.

?

a.?????Understanding which scenarios are suitable for AI implementation today

b.????Knowing trends in AI and transferring that knowledge in your organization

c.?????Laying right foundations for AI readiness for a large-scale transformation in next 5 to 10 years

d.????And preparing an enterprise for its ability to scale AI across all business functions will be key to personal career development

?

This is a second reason why doing this course will be helpful for your own career.

???????

3.????Job Security/ relevance:

There is a real threat to some job types but for majority of other jobs, their profiles will change that is, the collection of activities that form a job profile. The challenge that people and governments face today is the speed with which the change is happening and the scale of change.

?

?

AI will displace majority of jobs. Note I am not using the word “Replace” here although there is no denying at individual level, AI will also replace lots of jobs until an individual has shifted to new industry or gained relevant skills for the same job. I am of the firm belief that AI will also create new types of job which will not fit the definition of even the traditional IT jobs which have evolved in the last 20 years. One way or the other, workforce of today and tomorrow will have to learn these technologies in varying degrees of depth.

?

Countries with high R&D spend, mature VC market and supportive governments where new startups are encouraged are already running short of people with Digital and Machine learning skills. Private corporations are constantly replacing leaders and workforce with people who bring these new skills and are able to bring about the change that they see most value coming out from.

?

Doing a Machine Learning course will enable you

a.?????to use the relevant ML vocabulary

b.????to translate the domain requirements into a language ML developers can understand,

c.?????to identifying potential areas within your domain that are good candidates for ML applications,

d.????being able to sell these technologies to your customer by demonstrating the value of ML in your solution

e.????identify and develop AI training for all employees of your team or organization

?

These are a just fraction of reasons why knowledge of AI will not only make you relevant but valuable in the eyes of employers.

?

If you are a leader, there is all the more reason why you should be able to understand where in your business AI can bring value. How to differentiate one AI projects from another to align with organizational goals and how to prepare the organization for adopting ML at scale.

?

I have titled this section “Job security” which at least in my mind brings out negative feelings of fear and insecurity. I am however also indicating that knowing upcoming changes around us, risks or threats those changes may carry and preparing for it is a sign of prudence and wisdom. This article and especially this section is to help people think about waves of changes happening around us in the workplaces and to do something about their own careers. Choose to stay relevant. ???

?

?

4.????Career Transition

If you are someone looking for transitioning to a new career in the field of data Analytics, Machine learning / AI and related field, you need to do this course. ML Specialization is a first step to mastering your fundamentals and will also help you firm up your decision without making expensive mistakes later on.????

?

?

Barrier to completing Machine Learning Specialization

?

The next part of this article focuses on removing any barriers to doing this course itself.

?

While speaking to some of my colleagues, I realized people understand the relevance of learning about AI and its impact on their careers but are discouraged because of the fear of not being able to complete this course. And the reason for that fear is it is assumed that people need some form of coding mastery and deep knowledge of Math. Before delving in depth on this topic, let me make it absolutely clear:

?

To get value out of this course:

1.????You really do not need a coding background

2.????And you also do not really need mastery of Math

?

The instructor has designed this course for a wide range of audience but with a common purpose – to gently introduce the various concepts of Machine learning intuitively. ?

The Machine learning specialization on Coursera gently introduces the concepts of ML technology, various algorithms, techniques, the intuition (the how and why from a layman’s perspective), the Math behind it, where a ML model is applied, how it is applied, challenges and ways to improve the chances of ML project success.?

?

This specialization opens up your mind to understanding not only the AI skills but also to your own interests in how you want to design your future skills. Refer figure above – After this course, the biggest takeaway for anyone is the clarity on this subject. Secondly, you will know which direction you want to proceed:

?

1.????For example, you can leave at it for time being and use this knowledge to apply to role and domain expertise

2.????Become an ML evangelist (non-coding): Pursue more such courses and learn other ML applications from research papers, AI publications and podcasts. Use that knowledge to cross-fertilize ideas in your own domain. Publish, share your knowledge

3.????Take the ML engineer approach: Make a multi-year plan, figure out the gaps where you may need brushing up your foundations of math and coding. Learn TensorFlow and other ML libraries, do tons of projects and most importantly publish your learning

4.????The approach for students interested in AI career or people wanting to become Researcher should also follow the ML engineer approach

?

?

Like others, before doing this course, I felt I needed to get my math skills in place so I spent several hours learning statistics, linear Algebra and differential equations. I got bored and was immediately put off. It also set me back several times in my quest to learn AI. My suggestion, don’t spend any of your time on Math or coding until you do this course. Once you know where your interest is and if it includes becoming a researcher or ML engineer, you can then decide to fill the gaps in your math skills gradually, same goes for coding.

?

Let’s get a bit more into the course design and structure. It’ll reinforce my arguments around math and coding.

?

The Machine Learning Specialization is now in its second avatar. Following are some of the solid reasons to pursue this course:

?

1.????Relevance: I’ll say the most important characteristic of this specialization is the relevance to the current AI/ML trends in the market

2.????The specialization is extremely well structured: This Specialization has 3 courses; each course load is distributed across weeks and each week has 2-3 hours of videos averaging 8-10 mins in duration. Andrew takes through a concept, explains the math, the intuition and then demonstrates how the model is developed in a code. There are questions and exercises (Practice labs) after explanation of a concept. I’ll go through the exercises in a separate section below.?

3.????There is no need to download any software. The application used for Practice lab is called Jupyter Notebook which is available on the browser itself. Jupyter notebooks are extremely versatile and easy to use application. On this specialization, the only command that you have remember is to press the Run command button on the top ribbon or use the keyboard shortcut “Shift + Enter”. Easy-peasy ?????

4.????Student Support: The Deeplearning.AI has a Discord community which each student is encouraged to join. The access is available to the paid students of this specialization. It is a like a database of questions and answers in addition to being a Q&A forum. Chances are a student may already have asked the question that you want to ask and so in majority of cases a quick search will take you to an explanation. The discussion is also organized by Specialization – Course – Week – Topic format and so you know there is a direct relevance to a video or exercise. If you do not find your question already, just post the question and a technical mentor or someone with the understanding in the student community will answer in less than a few hours. If I remember correctly, you can also form your own support or study group if you enjoy studying in groups.

?

?

I am sure you are still thinking about the coding exercises or Practice labs. Let me go through this final hurdle that may be blocking you from taking this course.

?

?

Because this course was also designed for people with non-coding background, a number of practice labs are optional. My suggestion is to not skip it at all. Here are 3 reasons:

?

1.????The Practice lab summarizes the topic so it is like a final revision of the topic

2.????It is in the practice labs that you see how the math is turned into codes. The instructors explain the logic in plain English the intuition behind developing codes from concept. For people only interested in the Machine Learning concept, these intersections are a great way to understand what level of effort goes into turning any concept into a real code. Students or professional looking forward to becoming ML specialist will find these practice labs introduce different types of libraries and challenges in developing and deploying these codes.

3.????The practice labs will expose you to plethora of ML vocabulary especially useful to folks who are already or will interact with colleagues from Tech background

?

Most practice labs have lots of visual explanations, stuff to trial out and observe and a few questions. Often, these questions are the easiest part of the overall code. And there is a successive level of support available.

?

1.????Before asking students to attempt a question, the instructor explains in plain English what needs doing and even how should a student attempt to answer it

2.????And then there are 2-3 levels of hint provided for each question. Remember I mentioned earlier, this course was designed for a wide range of student which could include people without any coding background or not interested in pursuing a coding career. The hints provide sufficient level of information to complete the questions

?

?

Duration of specialization and cost

?

I’d say an hour/ week-day and 2-3 hrs/day of weekend over 3 months is a good estimate of time commitment required to complete this course. It took me 2.5 months with a week off in between and several days where I studied longer and certain days with just 15-20 mins max. The lesson I took away was – be consistent. Do not let gaps come up in between, spend at least 15 mins on days you cannot think it is possible for you to focus or dedicate. ?

?

People with more time to hand and students can certainly wrap it up faster.

?

Coursera charges a monthly fee for this specialization. I’d guess it is different in different countries. I think I paid around ￡39 over 3 months for this course. I’ll say it is worth more value than I paid.

?

?

Keeping abreast on AI

?

Although the YouTube is full of instructional videos and some of the instructors are great but YouTube is best to clarify or to observe an alternate explanation of a topic at best. So, for people looking to build their career in AI/ ML or wanting to keep themselves abreast of applications of AI, here is a small list of resources to start with:

??

1.????For clarity on courses to build an AI/ML career, I think this is the best starting point: ?TensorFlow ML Learning resources page:???

?

It is a carefully curated page of all the resources – free and paid that are considered best in industry to embark on career in AI/ ML

?

2.????I'd also suggest visiting the DeepLearning.AI Courses tab. The courses are organized by the stage of learning you are at and also includes some less spoken about courses like AI for medicine. Clicking on any these courses will take you to Coursera, the educational platform where these courses are actually delivered. ????

?

3.????Podcasts: I strongly recommend listening or watching the Deepmind podcasts by Hannah Fry, on YouTube or at Google podcasts. These are free and the reason they are great is because they have presented the advancement in AI in a fun, jargon-free and casual manner.

?

4.????For those who prefer reading over listening, check out Deepmind’s blog and I’d suggest simply spending time on their website to understand how AI is being used in different ways and most importantly the challenges in making the AI to work. I hope it will motivate you to think about the challenges in your own industry and one day possibly find a solution to a big problem.

?

?

??

I sincerely hope this article has helped you overcome any barriers in pursuing mastery of Machine learning. To summarize:

?

1.????Job profiles are changing and will continue to change significantly in coming decades

2.????There is a greater need for enhancing domain skills with digital skills especially knowledge of AI and how it affects your domain

3.????There is and will be a huge need for AI developers and Practitioners of all sort and so knowledge of AI will greatly enhance your relevance in marketplace

4.????It is critical to know how AI works, where can it be applied and how to seek value from it

5.????For people in leadership role and responsible for organizational transformation, it is equally important to know how to prepare your organizations for scaling AI, how to lay foundations for organization wide AI adoption. Taking this specialization is the first step to approach AI scaling

6.????Barriers to taking this or similar courses – Math and coding. How to overcome these barriers

7.????And finally, we went over the Coursera Machine Learning Specialization, how it is structured, Practice labs and planning for effort and cost.????

Thanks for taking the time out to read this article. If you have suggestions and resources to learn Machine learning that you would also like to share, please add it to comments.

?HAPPY LEARNING!!

Disclaimer: This article contains my personal views and has not been endorsed by nor have I been paid to promote the websites and education providers.