How to Build a Career in Artificial Intelligence. Tips from Andrew Ng

“AI is the new electricity. It will transform and improve all spheres of human life” (c) Andrew Ng

Read about the three key components of this scheme — learning basic skills, working on projects, and job hunting — below.

Coding AI — the new literacy

Today we take it for granted that many people can read and write. Someday, I hope, it will be just as common to be able to write code, particularly for AI.

A few hundred years ago, society did not consider language literacy a necessary skill. A small number of people could read and write, and everyone else simply did not interfere with them doing so. It took centuries for literacy to spread and enrich society.

Words provide deep communication between people. Code is the deepest form of communication between people and machines. As machines take up more and more space in everyday life, this communication becomes increasingly important.

Traditional software engineering — writing programs that explicitly tell the computer the sequence of steps to perform — has been the main path to code literacy. Many introductory programming courses use creating a video game or a website as an example. But AI, machine learning, and data science offer a new paradigm in which computers extract knowledge from data. This technology offers an even more efficient path to coding.

On Sundays, I buy a slice of pizza at the nearby pizzeria. The gentleman behind the counter has no particular reason to learn to create video games or write his own software for websites (besides personal growth and the pleasure of acquiring a new skill).

But AI and data science are significant even for a pizzeria. A linear regression model can better estimate demand to optimize the restaurant's staff and supply chain. The business will be able to better forecast sales of Hawaiian pizza — my favorite! — to prepare more in advance and reduce customer wait times.

The application of AI and data science is possible in almost any task related to obtaining data. Thus, various professions will find more applications for user AI applications and data analysis than for traditional software engineering. This makes literacy in AI-oriented coding even more valuable than traditional coding. It can help enrich the lives of countless people through data.

I hope that the prospect of creating basic AI applications, even more so than the prospect of creating basic traditional software, will encourage more people to learn to code. If society accepts this new form of literacy just as it does the ability to read and write, we will all benefit.

CHAPTER 1. Three Steps to Career Growth

The three key stages of career growth are learning basic skills, working on projects (to deepen skills, create a portfolio, and make an impact), and job searching. These stages overlap:

• Learning. Initially, you will focus on mastering basic skills.

• Projects. Once you have acquired basic technical skills, you will start working on projects. During this period, you will also continue to learn.

• Job. Later, you will start job searching. Throughout this process, you will continue to learn and work on meaningful projects.

These stages apply to various professions, but the AI field has some specifics. For example:

• Training. Mastering basic skills is a process that lasts throughout your career: AI is just emerging, and many technologies are still evolving. The basics of ML and deep learning have already been established, but keeping up with changing technologies in the AI field is more important than in more mature areas.

• Projects. Working on projects often involves collaborating with stakeholders who do not have sufficient experience in AI. This can complicate finding the right project, assessing its timelines and payback, and setting expectations. Additionally, the iterative nature of AI projects leads to specific challenges in project management: how to develop a plan for creating a system when it is not known in advance how long it will take to achieve the desired accuracy? Even after the system has reached its goal, additional iteration may be required to eliminate drift after deployment.

• Work. Here, one of the nuances is the inconsistency of opinions about AI skills and job roles. While finding a job in AI may be similar to finding a job in other industries, there are important differences. Many companies are still trying to figure out what AI skills they need and how to hire people who have them. What you worked on may be significantly different from what your interviewer has seen, and you will most likely have to explain some aspects of your work to potential employers.

At each stage, you should also ensure that you have a supportive community around you. Having friends and allies who can help—and whom you strive to help—makes the journey easier. This is true whether you are taking your first steps or have been on this path for many years.

CHAPTER 2. Learning technical skills for a promising career in AI

Above, I described three key steps to building a career in AI, and all of these steps should be supported by community involvement. In this chapter, I will focus more on the first step: learning basic skills.

More scientific papers have been published on AI than anyone can read in a lifetime. Therefore, it is very important to set the right priorities when choosing a topic during training. I believe that the most important topics for a technical career in machine learning are:

• Fundamental ML skills. For example, it is important to understand models such as linear regression, logistic regression, neural networks, decision trees, clustering, and anomaly detection. Beyond specific models, it is even more important to understand the core concepts underlying how and why ML works, such as bias/variance, loss functions, regularization, optimization algorithms, and error analysis.

• Deep learning. This field has become so extensive that it is difficult to succeed in ML without understanding it! It is useful to know the basics of neural networks, the intricacies of their tuning (such as hyperparameter tuning), convolutional neural networks, recurrent networks, and transformers.

• Mathematics in the context of ML. Key areas include linear algebra (vectors, matrices, and various manipulations with them), as well as probability and statistics (including discrete and continuous probability, standard probability distributions, basic rules such as independence and Bayes' rule, and hypothesis testing). Additionally, exploratory data analysis (EDA) — using visualizations and other methods to systematically explore a dataset — is an underrated skill. I find EDA particularly useful when developing data-centric AI, where error analysis and insights can really drive progress. Finally, a basic intuitive understanding of calculus also helps. The mathematics needed for effective machine learning is constantly changing. For example, while some tasks require calculus, advanced automatic differentiation software allows for the invention and implementation of new neural network architectures without any calculus. Ten years ago, this was practically impossible.

• Software development. While you can get a job and make a huge contribution with just ML modeling skills, your chances of employment will increase if you can also write good software to implement complex AI systems. This requires familiarity with programming basics, data structures (especially those related to ML, such as data frames), algorithms (including those related to databases and data manipulation), software design, Python, and key libraries such as TensorFlow or PyTorch, as well as scikit-learn.

• This needs to be learned! Even after you master everything on this list, I hope you will continue to learn and deepen your technical knowledge. I know many ML engineers who have benefited from deeper skills in applied areas such as natural language processing or computer vision, or in technological areas such as probabilistic graphical models or building scalable software systems.

• How to acquire these skills? There is a lot of good content online, and theoretically reading dozens of web pages might work. But when the goal is deep understanding, reading scattered web pages is inefficient because they repeat each other, use inconsistent terminology (which slows down the process), vary in quality, and leave gaps. This is why a good course, where the material is organized into a coherent and logical form, is often the most effective way to master a significant amount of knowledge. Once you have mastered the knowledge available in courses, you can move on to research papers and other resources.

Finally, no one can learn everything they need to over the weekend or even in a month. All the people I know who have succeeded in ML study all their lives. Given how quickly our field is changing, you have no choice but to keep learning if you want to keep up with the times.

How to maintain a steady pace of learning over the years? If you can develop the habit of learning a little each week, you can make significant progress with seemingly less effort.

The best way to develop a new habit

One of my favorite books is "Tiny Habits: Small Changes That Change Everything" by BJ Fogg. Fogg explains that the best way to develop a new habit is to start small and succeed, rather than start big and fail. For example, instead of trying to exercise for 30 minutes a day, he recommends aiming to do just one push-up and doing it consistently.

This approach can be helpful for those who want to spend more time studying. If you start by watching, say, 10 seconds of an educational video every day — and do it consistently — the habit of studying daily will form naturally. Even if you don't learn anything in those 10 seconds, you will develop the habit of learning a little every day. On some days, you may study for an hour or longer.

CHAPTER 3. Do you need to study math to get a job in AI?

How deeply do you need to know math to become an ML engineer?

Is math a fundamental skill for AI? It's always nice to know more math! But this science is so vast that in reality, you have to prioritize. Here's how you can strengthen your math knowledge.

To begin with, I find it helpful to ask: what do you need to know to make the decisions necessary for the job you want to do? At DeepLearning.AI, we often ask: "What does a person need to know to achieve their goals?" The goal might be building an ML model, creating a system architecture, or passing an interview.

Understanding the mathematics underlying the algorithms used is often useful because it allows you to debug them. But the depth of useful knowledge changes over time. As ML methods evolve and become more reliable, they require less debugging, and a more superficial understanding of mathematics may be sufficient for their operation.

For example, in an earlier era of ML, linear algebra libraries for solving linear systems of equations (for linear regression) were immature. I needed to understand how these libraries worked to choose between different libraries and avoid pitfalls associated with rounding numbers. But this has become less important as numerical linear algebra libraries have evolved.

Deep learning is still an evolving technology, so when you train a neural network and the optimization algorithm cannot converge, understanding the mathematics underlying gradient descent, momentum, and the optimization algorithm Adam will help you make better decisions. Similarly, if your neural network does something funny—say, poorly predicts images of a certain resolution but not others—understanding the mathematics underlying neural network architecture will help you better understand what to do.

Of course, I also encourage curiosity-driven learning. If something interests you, go and study it, regardless of how useful it may turn out to be! It may lead to a creative spark or a technical breakthrough.

CHAPTER 4. Defining the Scope of Successful AI Projects

One of the most important skills of an AI architect is the ability to find ideas worth working on. In the next few chapters, we will discuss finding and working on projects so that you can gain experience and build your portfolio.

Over the years, I have had the opportunity to apply ML in manufacturing, healthcare, climate change, agriculture, e-commerce, advertising, and other industries. As a person who is not an expert in all these industries, how can one find meaningful projects in them? Here are five steps to help you determine the scope of projects.

1. Identify the business problem (not the AI problem). I like to find an expert in the field and ask: "What are the three things you would like to improve? Why haven't they been improved yet?" For example, if you want to apply AI to combat climate change, you may find that grid operators cannot accurately predict how much energy intermittent sources like wind and sun can generate in the future.

2. Brainstorm IT solutions. When I was younger, I used the first idea that came to mind. Sometimes it worked, but sometimes I missed an even better idea that would not have taken more effort to implement. Once you understand the problem, you can more effectively brainstorm potential solutions. For example, to predict power generation from intermittent sources, you can use satellite images to more accurately determine the location of wind turbines, assess their height and power, or weather data to better predict cloud cover and, consequently, solar radiation. Sometimes a good AI solution does not exist, and that's okay too.

3. Evaluate the feasibility and value of potential solutions. You can determine the technical feasibility of an approach by reviewing published work, competitors' experience, or perhaps creating a quick proof-of-concept implementation. You can determine its value by consulting with experts in the field (e.g., grid operators who can provide recommendations on the usefulness of the potential solutions mentioned above).

4. Define the stages. Once you consider the project valuable enough, the next step is to define the metrics to aim for. These include both ML metrics (e.g., accuracy) and business metrics (e.g., revenue). ML teams often find it most convenient to work with metrics that the training algorithm can optimize. But we may need to step out of our comfort zone to come up with business metrics, such as those related to user engagement, revenue, and so on. Unfortunately, not every business problem can be reduced to optimizing test set accuracy! If you can't come up with reasonable benchmarks, it may be a sign that you need to learn more about the problem. A quick proof of concept can help gather the missing information.

5. Budget for resources. Think through everything you will need to implement the project, including data, personnel, time, as well as any integrations or support that may be required from other teams. For example, if you need funds to purchase satellite images, make sure this is included in the budget.

Working on projects is an iterative process. If at any stage you find that the current direction is not feasible, go back to the previous stage and continue working with the new understanding. Is there an area that interests you and where AI can be beneficial? I hope these steps help you explore it within the framework of project work — even if you don't yet have deep knowledge in this area. AI won't solve all problems, but as a community, let's look for ways to make a positive impact wherever possible.

CHAPTER 5. Finding projects that align with your career goals

Of course, we should only work on projects that are ethical and beneficial to people. But these constraints leave a wide choice. In the previous chapter, I wrote about how to determine the scale of AI projects. In this and the next chapters, the focus is somewhat different: choosing and executing projects with a focus on career growth.

A fruitful career will include many projects that, we hope, will become increasingly large, complex, and influential over time. Thus, starting small is perfectly fine. Use the first projects for learning and gradually move on to larger projects as your skills grow.

When you are just starting out, don't expect great ideas or resources to be handed to you on a silver platter. Many people start with small projects in their spare time. Once you have your first successes — even small ones — your growing skills increase your ability to propose better ideas, and it becomes easier to convince others to help you move on to larger projects.

What to do if you have no project ideas? Here are some ways to generate them:

• Join existing projects. If you find someone else with an idea, ask to join their project.

• Keep reading and talking to people. I get new ideas when I spend a lot of time reading, attending courses, or talking to experts in the field. I'm sure you can too.

• Focus on the application area. Many researchers try to advance basic AI technologies — say, invent the next generation of transformers or further expand language models — and while this is an exciting direction, it is also very difficult. But the variety of applications where ML has not yet been applied is huge! I was lucky that I was able to apply neural networks to everything from autonomous helicopter flight to online advertising, partly because I got into the game when relatively few people were working on these applications. If your company or educational institution is interested in a specific application, explore the possibilities of ML. This can allow you to take a first look at a potentially creative application where you can do unique work that no one has done before.

• Start a side business. Even if you have a full-time job, an exciting project that may or may not grow into something more can stir up creative juices and strengthen connections with colleagues. When I worked as a full-time professor, working on online education was not part of my "job" (which was to do research and teach classes). It was a fun hobby that I often worked on out of a passion for education. My early experience recording videos at home helped me later work on online education more thoroughly. Silicon Valley is full of stories of startups that started as side projects. If it doesn't cause a conflict with your employer, such projects can be a stepping stone to something significant.

If you have several project ideas, which one should you start?

Here is a brief list of factors to consider:

• Will the project help your technical growth? Ideally, it should be challenging enough to develop your skills, but not so difficult that you have little chance of success. This will set you on the path to mastering increasingly complex technical tasks.

• Do you have colleagues at work and people in general with whom you can discuss the project? We learn a lot from the people around us, and good companions will have a huge impact on your growth.

• Can the project become a stepping stone? If it turns out to be successful, will its technical complexity and/or impact on the business make it a significant step towards larger projects? If the project is larger than those you have worked on before, it is likely that it can become such a stepping stone.

Finally, avoid analysis paralysis. There is no point in spending a month deciding whether to work on a project that will take a week. Throughout your career, you will work on many projects, so you will have plenty of opportunities to refine what exactly is worth doing. Given the vast number of possible projects in AI, instead of the usual "ready, aim, fire" approach, you can speed up your progress with the "ready, fire, aim" approach.

Ready, fire, aim

Working on projects requires making complex decisions about what to create and how to do it. Here are two different styles:

• Ready, aim, fire: carefully plan and conduct thorough verification. Commit and start executing only when you have a high degree of confidence in the chosen direction.

• Ready, fire, aim: start developing and begin executing right away. This will allow you to quickly identify problems and make changes if necessary.

Suppose you have created a chatbot for customer service in retail and think it can also help restaurants. Should you spend time studying the restaurant market before starting development, moving slowly but reducing the risk of wasting time and resources? Or should you jump right into action, moving quickly and taking on the increased risk of pivoting or failing?

Both approaches have their proponents, and the best choice depends on the situation.

"Ready, aim, fire" is generally better when the cost of execution is high and research can shed light on how useful or valuable the project might be. For example, if you can brainstorm several other use cases (restaurants, airlines, telecommunications companies, etc.) and evaluate them to determine the most promising one, it may be worth spending extra time before choosing a direction.

"Ready, fire, aim" is generally better if you can execute the project at low cost, while determining if the direction is feasible and finding changes that will make it work. For example, if you can quickly create a prototype to understand if the product is needed by users, and if cancellation or pivoting after a small amount of work is acceptable, then it makes sense to jump into action quickly. When the shot is cheap, it makes sense to take many shots. In this case, the process looks like "Ready, fire, aim, fire, aim, fire, aim, fire, aim, fire".

After agreeing on the project direction, when it comes to building an ML model that will be part of the product, I lean towards "Ready, fire, aim". Building models is an iterative process. For many applications, the cost of training and error analysis is not prohibitive. Additionally, it is very difficult to conduct research that will shed light on the appropriate model, data, and hyperparameters. Therefore, it makes sense to quickly create an end-to-end system and refine it until it works well.

But if choosing a direction means costly investments or entering a door with one-way traffic (i.e., a decision that is difficult to reverse), it is often worth spending more time to make sure it is indeed a good idea.

CHAPTER 6. Creating a portfolio of projects demonstrating skill growth

Throughout your career, you are likely to work on successive projects that will become increasingly large and complex. For example:

1. Educational projects. The first few projects may be narrowly focused assignments with predetermined correct answers. Often, this is a great learning experience!

2. Personal projects. You can continue working on small projects alone or with friends. For example, reworking a known algorithm, applying ML to your hobby (predicting whether your favorite sports team will win, etc.), or creating a small but useful system at work in your spare time (e.g., an ML-based script to help a colleague automate part of their work). Participating in competitions like those organized by Kaggle is also one way to gain experience.

3. Creating value. Over time, you will acquire enough skills to create projects that others see tangible value in. This will give you access to greater resources. For example, instead of developing ML systems in your spare time, you can make it part of your job, gaining access to more equipment, computing time, labeling budget, or staff.

4. Increasing scale and complexity. Successes follow one another, paving the way for technical growth, increased resources, and increasingly significant projects.

Each project is just one step on a long journey that will hopefully be fruitful. Additionally:

• Don't be afraid to start small. One of my first ML research projects involved training a neural network to see how well it could mimic the sin(x) function. It wasn't very useful, but it was a great experience that allowed me to move on to larger projects.

• Communication is key. You need to be able to explain your thoughts if you want others to see the value of your work and trust you with the resources you can invest in larger projects. At the start of a project, explaining its potential value will help attract colleagues, mentors, and managers—and point out flaws in your reasoning. After completing the project, being able to clearly explain what you have achieved will help convince others to open the door for you to larger projects.

• Leadership is not just for managers. When you get to work on large AI projects that require teamwork, your ability to lead projects will become more apparent, whether you hold a formal leadership position or not. Many of my friends have successfully pursued technical rather than managerial careers, and their ability to help lead a project by applying deep technical knowledge—such as when to invest in a new technical architecture or collect more data of a certain type—has allowed them to grow as leaders and significantly improve the project.

Building a portfolio of projects, especially those that show progress over time from simple to complex, will be a great help when looking for a job.

CHAPTER 7. A Simple Framework for Starting a Job Search in AI

The job search consists of several predictable stages, including choosing the companies you want to apply to, preparing for interviews, and finally choosing a role and negotiating salary and bonuses. In this chapter, I would like to focus on a framework that will be useful to many AI job seekers, especially those coming from another field.

When thinking about your next job, ask yourself:

• Are you changing roles? For example, if you are a software engineer, university student, or physicist who wants to become an ML engineer, this is a role change.

• Are you changing industries? For example, if you work in a healthcare company, financial company, or government agency and want to work in a software development company, this is an industry change.

A product manager in a tech startup who becomes a data scientist in the same company (or another) has changed roles. A marketer in a manufacturing company who becomes a marketer in a tech company has changed industries. An analyst in a financial company who becomes an ML engineer in a tech company has changed both roles and industries.

If you are looking for your first job in AI, it will probably be easier for you to change roles or industries than to do both at the same time. Suppose you are an analyst working in financial services:

• If you find a job in data science or ML in financial services, you can continue to use your domain knowledge while gaining knowledge and experience in AI. After working in this position for some time, you will have more opportunities to transition to a tech company (if that is still your goal).

• Additionally, if you become an analyst in a tech company, you can continue to use your analytical skills but apply them in a different industry. In a tech company, it is also much easier to learn from colleagues about practical AI problems, key skills needed for success in AI, and so on.

If you are considering a role change, it is easier to do so in a startup than in a large company. Although there are exceptions, startups usually lack the people to do all the necessary work. If you can help with AI tasks — even if it is not your official job — your work is likely to be appreciated. This lays the foundation for a possible role change without having to leave the company. While in a large company you are more likely to be rewarded for doing your job well (and your manager for supporting you in doing the job you were hired for), you are less likely to be rewarded for contributions beyond it.

After working for some time in the desired position and industry (for example, as an ML engineer in a technology company), you will have a good understanding of the requirements for this position in this industry at a higher level. In addition, you will have industry contacts that will help you. Therefore, future job searches — if you decide to stay in this position and in this industry — are likely to be easier.

When changing jobs, you take a step into the unknown, especially if you are changing roles or industries. One of the most underutilized tools for getting to know a new role and/or industry is the informational interview. I will talk more about this in the next chapter.

I am grateful to Salva Nur Muhammad, CEO of FourthBrain (a branch of DeepLearning.AI), for providing some of the ideas outlined in this chapter.

Overcoming Uncertainty

We don't know much about the future. When will we cure Alzheimer's disease? Who will win the next election? Or, in the context of business, how many customers will we have next year?

With so many changes happening in the world, many people feel stressed about the future, especially when it comes to job hunting. I have a practice that helps me regain a sense of control. When faced with uncertainty, I try to:

1. Make a list of likely scenarios, acknowledging that I don't know which ones will come true

2. Make an action plan for each scenario

3. Start taking actions that seem reasonable

4. Periodically review scenarios and plans as the future approaches

For example, during the Covid-19 pandemic in March 2020, I conducted a scenario planning exercise. I envisioned a quick (three months), medium (one year), and slow (two years) recovery from Covid-19 and made plans to manage each case. These plans helped me prioritize.

This same method can be applied to personal life. If you are unsure whether you will pass an exam, get a job offer, or a visa — all of which can be stressful — you can write down what you will do in each of the likely scenarios. Thinking through possibilities and executing plans will help you navigate the future effectively, whatever it may bring.

Bonus: by learning AI and statistics, you can calculate the probability of each scenario. I am a fan of the superforecasting methodology, where the judgments of many experts are synthesized into a probability estimate.

CHAPTER 8. Using Informational Interviewing to Find the Right Job

If you are preparing for a role change (e.g., starting a job as an ML engineer for the first time) or industry change (e.g., joining a company specializing in AI technologies for the first time), you probably don't know much about your future workplace. The technique known as informational interviewing is a great way to learn about it.

Informational interviewing involves finding someone in a company or position you would like to learn more about and having an informal interview with them about their job. These conversations are not directly related to job searching. Moreover, it is helpful to talk to people in positions that match your interests long before you are ready to start job hunting.

• Informational interviews are especially relevant for AI. As this field evolves, many companies use job titles inconsistently. In one company, data analysts may be expected to primarily analyze business data and present findings in the form of slides. In another company, they may write and maintain production code. An informational interview will help you find out what AI specialists actually do in a particular company.

• With the rapid expansion of AI capabilities, many people are entering this field for the first time. In this case, an informational interview can be invaluable in obtaining information about what is happening and what skills are needed to perform the job well. For example, you can learn about the algorithms, deployment processes, and software stacks used by a particular company. You might be surprised, if you are not yet familiar with the data-driven AI movement, to learn how much time most ML engineers spend on iterative data cleaning.

When preparing for informational interviews, research the interviewer and the company in advance to come up with thoughtful questions. You can ask:

• What do you do on a typical week or day?

• What tasks are most important in this role?

• What skills are most important for success?

• How does your team work together to achieve its goals?

• What is the hiring process like?

• When considering candidates who have stood out in the past, what made them shine?

Finding someone to interview is not always easy, but many people in high positions today received help early in their careers from those who entered the field before them, and many are eager to repay the favor. If you can reach out to someone already in your network—perhaps a friend who got a job before you, or someone who went to the same school as you—that's great! Events like Pie & AI can also help you find useful connections.

Finally, be polite and professional and thank the people you interviewed with. And when you have the opportunity, help others with information yourself. If you receive a request for an informational interview from someone in the DeepLearning.AI community, I hope you agree to help them take a step forward! If you are interested in learning more about informational interviews, I recommend this article from the UC Berkeley Career Center.

I have mentioned several times the importance of the people and community around you. People not only provide useful information, but can also play an invaluable role in directing you to potential employers.

CHAPTER 9. Finding the Right Job in AI

In this chapter, I would like to discuss some of the nuances of job searching.

A typical job search follows a fairly predictable path.

• Research roles and companies online or talk to friends.

• Alternatively, arrange informal informational interviews with employees of companies you are interested in.

• Either contact companies directly or, if you can, get a referral from someone on staff.

• Interview with companies that send you an invitation.

• Get one or more offers and choose one. Or, if you did not receive an offer, ask for feedback from interviewers, HR staff, online discussion forums, or anyone else who can help you plan your next step.

The process may look familiar, but each job search is different from the other. Here are some tips to find a position that will contribute to your successful career and allow you to grow further.

• Pay attention to the basics. A compelling resume, a portfolio of technical projects, and a strong interview performance will open doors for you. Even if you received a recommendation from someone within the company, your resume and portfolio will be your first contact with many people who do not yet know about you. Update your resume and make sure it clearly presents your education and experience relevant to the desired role. When communicating with each company, explain why you are a good fit for them. Before the interview, ask the recruiter what is expected of you. Take the time to review and practice answers to common interview questions, hone key skills, and study technical materials. After the interview, take notes to remember what was said.

• Act respectfully and responsibly. Approach interviews and offer negotiations with a winning attitude. On social media, outrage spreads faster than reasoned arguments, so the story of an employer underpaying someone becomes louder, while the story of an employer treating someone fairly does not. The vast majority of employers are ethical and fair, so do not let stories about a small number of people being mistreated affect your approach. If you leave a job, leave gracefully. Give your employer advance notice, give your best effort in the last hour of work, hand over unfinished business as best as possible, and leave in a way that honors the responsibilities entrusted to you.

• Choose who to work with. The colleagues you work with are just as important as the projects themselves, as we are influenced by the people around us. For example, if your friends smoke, the likelihood that you will also smoke increases. I do not know of a study that shows this, but I am almost certain that if most of your colleagues work hard, constantly learn, and create AI for the benefit of all people, you are likely to do the same.

• Seek help from your surroundings. Most of us go job hunting only a few times in our careers, so few people get much practice in doing it well. However, collectively, the people in your immediate surroundings surely have a lot of experience. Do not hesitate to reach out to them. Friends and colleagues can give advice, share knowledge, and direct you to other people who can help. When I applied for my first teaching position, friends and mentors helped me a lot, and many of the tips they gave me proved to be extremely useful.

I know that the job search process can be daunting. Instead of viewing it as a big leap, consider a gradual approach. Start by identifying potential roles and conducting a few informational interviews. If as a result of these conversations you realize that you still have a lot to learn before you are ready to send out your resume, that's great! At least you have a clear path forward. The most important part of any journey is to take the first step, and that step can be very small.

CHAPTER 10. Keys to Building a Career in AI

The path to career success in AI is much more complex than what I can describe in one short e-book. I hope the previous chapters will give you the momentum to move forward.

1. Teamwork. When we take on large projects, we achieve greater success by working as a team rather than individually. The ability to collaborate with others, influence them, and be influenced by them is crucial. Thus, interpersonal and communication skills are indeed very important. (By the way, I used to be a rather poor communicator).

2. Networking. I hate networking! For me, as an introvert, going to a party to smile and shake as many hands as possible is an activity bordering on horror. I would rather stay home and read a book. Nevertheless, I am fortunate that in the AI field I have found many true friends; people I would gladly go to battle for and who I also rely on. No person is an "island," and having a large circle of acquaintances can help you move forward when you need help or advice. Instead of creating "networks," I find it more useful to think about creating a community. So instead of trying to build my personal network, I focused on creating communities that I am part of. This helps me meet more people and make friends.

3. Job search. Of all the stages of career building, this one tends to attract the most attention. Unfortunately, there is a lot of bad advice on this subject on the Internet. (For example, many articles encourage taking a hostile stance towards potential employers, which I don't think is helpful). While it may seem that finding a job is the ultimate goal, it is just one small step on a long career path.

4. Personal discipline. Few people will know whether you spend your weekends studying or watching TV, but over time they will notice the difference. Many successful people develop good habits in nutrition, exercise, sleep, personal relationships, work, learning, and self-care. Such habits help them move forward while staying healthy.

5. Altruism. I believe that people who strive to help others at every step of their own path often achieve better results. How can we help others while building an exciting career for ourselves?

CHAPTER 11. Overcoming Impostor Syndrome

Before we move on to the final chapter of this book, I would like to address a serious issue that newcomers to the field of AI sometimes experience impostor syndrome. When someone — regardless of their achievements in this field — wonders if they are a fraud and truly belong to the AI community. I want to make sure that this does not deter you or anyone else from developing in the field of AI.

Let me be clear: if you want to be part of the AI community, I welcome you with open arms.

It is estimated that 70 percent of people experience impostor syndrome at some point. Many talented people have publicly talked about it, including US First Lady Michelle Obama, actor Tom Hanks, and Atlassian CEO Mike Cannon-Brookes. This happens in our community even among established people. If you have never experienced this yourself, that's great! I hope you will join me in encouraging and welcoming everyone who wants to join our community.

AI is technically complex, and there are a fair share of smart and very capable people in this field. But it is easy to forget that to become good at anything, the first step is to become a failure. If you have succeeded in lousy AI — congratulations, you are on the right track!

There was a time when I struggled to understand the math behind linear regression. I was puzzled when logistic regression worked strangely with my data, and it took me several days to find an error in my implementation of a basic neural network. Today I still find it difficult to read many scientific papers, and I recently made an obvious mistake when setting the hyperparameters of a neural network (fortunately, it was noticed and corrected by a fellow engineer).

So if you also find some parts of AI difficult, that's okay. We've all been there. I guarantee that everyone who has published a seminal work on AI has struggled with similar technical issues at some point.

Here are some things that might help.

• Do you have supportive mentors or peers? If not yet, attend Pie & AI or other events, use discussion forums, and try to find them. If your mentors or supervisor do not support your growth, find those who do. I am also working on developing a supportive AI community, and I hope that finding and providing support will become easier for everyone.

• No one is an expert in everything. Acknowledge what you are good at. If you understand and explain to your friends one-tenth of the articles in The Batch, you are on the right track! Let's work on understanding two-tenths of all the articles.

My three-year-old daughter (who can barely count to 12) regularly tries to teach my one-year-old son something. No matter how far you have come — if you are knowledgeable at the level of a three-year-old — you can encourage and uplift others who are behind you. This will also help you, as those who stand behind you will recognize your expertise and encourage you to further develop. When you invite others to join the AI community, which I hope you will, it will also reduce any doubts that you are already one of us.

AI is such an important part of our world that I would like everyone who wants to be a part of it to feel at home within our community. Let's work together to make this happen.

Final thoughts. Count every day

Every year on my birthday, I think about the days that have passed and those that may lie ahead.

You may be good at math; I am sure you can answer the following question with quick calculations. But please try to answer it based on your intuition, without calculations.

How many days does an average human life consist of?

When I ask friends, many name a number in the hundreds of thousands. (Many others, to my annoyance, cannot resist calculating the answer!)

I remember when I was a graduate student, I entered my statistical data into a mortality calculator to calculate my life expectancy. The calculator showed that I could expect 27,649 days of life. I was struck by how small that number was. I printed it out in large print and hung it on the wall of my office as a daily reminder.

These are all days that we can spend with loved ones, learn, build the future, and help others. Whatever you are doing today, is it worth 1/30,000 of your life?