AI For Everyone Course Review & Key Takeaways

I am working in ML/AI field for 6 years now and apart from technical skills that I acquired while working on the projects, I have also discussed various aspects of ML/AI with my non-technical colleagues, who have mostly been senior manager, VPs or CXOs.

When I heard about “AI For Everyone” course, I was a bit reluctant in attending it as I thought I know most of the generic stuff that might have been talked in the course.

Recently, one of my colleagues discussed with me a few topics covered in this course which intrigued me to get a fresh perspective on these topics. So, I recently attended this course on Coursera.

My motivation to write this blog is to make sure that I have understood key aspects of this course and am able to make my non-technical colleagues and project stakeholders understand the benefits & limitations of using AI.

This blog will also serve as a refresher for the professionals who have already attended it or are working in AI field but couldn’t attend the course due to time limitation.

Review of the Course

“AI For Everyone” is a short non-technical course which helps to bridge the gap among corporate professionals who are going to be impacted by AI transformation now or in the near future. Even the professionals without programming background understand the capabilities and limitations of today’s AI, what it takes to incorporate AI into your company’s strategy, how some of the fear regarding AI is overhyped and finally many serious questions like how AI will impact repetitive automation-prone jobs.

Key aspects of this course are:

• A non-technical overview of AI field
• Mainly for non-technical staff including executives (VPs & CXOs)
• Refresher for AI starters & enthusiasts
• Brings all stakeholders for AI projects on the same page
• Mostly covers benefits & limitations of AI

In my view, the most important benefit of this course is that it brings all the professionals on the same page about what is AI? What are its capabilities? Which really fills a lot of gaps for the people joining AI projects from different background and reduces the friction while working together.

Outline of the Course

Let’s have a look at what we are going to learn in this course:

• First week: AI technology, what is AI and what is machine learning? What’s supervised learning, that is learning inputs, outputs, or A to B mappings. As well as what is data science, and how data feeds into all of these technologies? What AI can and cannot do?
• Second week: What it feels like to build an AI project? What is the workflow of machine learning projects, of collecting data, building a system and deploying it, as well as the workflow of data science projects? How to carry out technical diligence to make sure a project is feasible, together with business diligence to make sure that the project is valuable before you commit to taking on a specific AI project?
• Third week: How such AI projects could fit in the context of our company? Examples of complex AI products, such as a smart speaker, a self-driving car. What are the roles and responsibilities of large AI teams? The AI transmission playbook, what are the five-steps for helping a company become a great AI company?
• Last week: AI and Society. What are the limitations of AI beyond just technical ones? How AI is affecting developing economies and jobs worldwide?

Week 1: What is AI?

This week’s content contains all the essentials for understanding AI, its capabilities, limitations, and how to implement it in a company’s automated processes to become an AI company.

Introduction

• AI is creating value in almost all the sectors
• Artificial Intelligence can be categorized broadly into two streams: ANI & AGI
• Artificial General Intelligence (AGI): Machine can do whatever a human can do
• Artificial Narrow Intelligence (ANI): Machine can do very specific tasks like driving cars, playing asked music, searching web etc
• What we know how to develop today are ANIs, each specialized in one task
• AGI doesn’t exist yet (at least for a long time from now)

Machine Learning

• AI has really taken off recently due to the rise of neural networks and deep learning
• In the majority of the cases: AI → Machine Learning → Supervised Learning
• Supervised Learning learns the mapping between input A to output B

What is data

• We know that Supervised Learning learns the mapping between input A to output B
• Examples of input A & output B from which ML algorithms learn is called data
• Data is often unique to your business, can be structured (tables, hierarchy etc), unstructured ( text, voice and image) or semi-structured
• Data can be obtained by: manual labeling, observing behaviours or download from websites/partnerships
• Pre-process the data before using it for AI models: garbage in, garbage out
• Most of the data problems are related to incorrect labels & missing values

The terminology of AI

• Business Intelligence (BI) helps interpret past data, BI is mainly used for reporting or Descriptive Analytics.
• Data Science is the science of extracting knowledge and insights from data which often ends up being captured in a deck.
• Machine Learning is the field of study that gives computers the ability to learn without being explicitly programmed.
• The Machine Learning which is a subset of the AI ​​itself contains subsets of which the most powerful is the Deep Learning (DL).
• Originally inspired by neural networks of the brain, DL models are composed of several layers of computational units (artificial neurons), each capable of detecting more and more complicated characteristics of a training dataset

What makes an AI company?

• Any company + deep learning != AI company
• Strategic data acquisition
• Unified data warehouse
• Pervasive automation
• New roles & division of labour

5 steps to becoming an AI company

• Execute pilot projects to gain momentum
• Build an in-house AI team
• Provide broad AI training
• Develop an AI strategy
• Develop internal & external communications

What Machine Learning can and cannot do

• ML/AI can’t do everything
• Expectations are inflated due to media, academics & only positive reports about AI
• Imperfect rule: Whatever we could do with a second of thought can probably be automated using AI
• Technical diligence is required to assess the feasibility of the project
• It is still harder to contextualize the responses for AI models
• ML tends to work well in learning a simple concept with lots of available data
• ML tends to work poorly in learning a complex concept, especially from small amount of data or if performing on new type of data

Intuitive explanation of deep learning

Week 2: Building AI Projects

The content of this week contains all the essentials for developing an AI project (large or small) and understanding how all jobs will be impacted by AI.

Workflow of a Machine Learning project

• Collect data
• Train model (iterate until its good enough)
• Deploy model (get data back, maintain/update model)

Workflow of a Data Science project

• Collect data
• Analyse data (iterate until its good enough)
• Suggest hypotheses/actions (deploy changes, re-analyse periodically)

Every job function needs to learn to use data

• Increasing sales: optimizing sales funnel & automated lead sorting
• Default detection: optimizing manufacturing line & automated visual inspection
• Recruitment assistance: optimizing recruitment funnel & automated resume screening
• Product recommendation: A/B testing & customized product recommendation
• Precision agriculture: crop analytics & precise weed killing

How to choose an AI project

• organize a cross-functional brainstorming with both the AI and domain teams
• Cross-section between ‘What AI can do’ & ‘What is valuable for your business’
• Which either increase revenue or reduce costs or both
• Use/buy standard AI models and develop only which don’t exist or offer extra business value

Method I

• Automation of processes/tasks rather than jobs
• Selection of activities with the greatest impact on the company’s business
• Selection of the main pain points in the company’s business

Method II

• Technical diligence in terms of state of the art AI capability, data requirements & time and people required to create, train & deploy AI models
• Business diligence in terms of cost reduction, revenue increment & new product/business opportunity
• Ethical considerations

Working with an AI team

• Instead of hiring AI specialist, train IT engineers in AI
• Domain team must define the performance rate to be achieved by the AI ​​model
• Not realistic to ask the AI ​​team to obtain a 100% performance rate
• Due to limitations of ML, insufficient data, mislabelled data

Technical tools for AI teams

• AI ​​evolves largely in the Open Source world
• Free ML/DL high-quality frameworks
• Academic research papers in ML/DL
• Refer Arxiv for research publications
• Refer GitHub for Open Source repositories
• AI ​​team must also have the significant computational capacity (GPU)
• Can be local or by prominent cloud providers (AWS, Azure & GCP)

Big data is not always required

• Having more data almost never hurts
• Data makes some businesses defensible
• But with small data, you can still make progress

Week 3: Building AI in your Company

This week’s content contains all the essential elements for developing AI in an organization (association, public organization, company).

Case study: Smart speaker

• Typical examples: Amazon Echo/Alexa, Google Home, Apple Siri, Baidu DuerOS
• Steps to process the voice command:

— Trigger word/wakeword detection

— Speech recognition

— Intent recognition

— Execute task

• Usually, there is one AI team per process
• Execution of the task may be more complex to be broken down in further sub-tasks
• Example tasks: play music, volume up/down, make call, current time etc

Case study: Self-driving car

• There are 3 main steps for a self-driving car to decide on its route and speed
• But there are many other localization processes that are necessary for the decision
• Key steps: Car detection → Pedestrian detection → Motion planning

Example roles of an AI team

• Software Engineer: 50% or more of the team, in charge of developing software
• ML Engineer: responsible for creating and training ML models
• ML Researcher: in charge of following the evolution of state-of-the-art, doing research and possibly publishing the results of its research
• Applied ML Scientist: in charge of adapting already published models to the specific projects of its company.
• Data Scientist: responsible for examining the data and providing the insights to AI team/executives
• Data Engineer: in charge of organizing and saving data in an accessible, secure and cost-effective way
• AI Product Manager: help AI team in deciding what to build, what’s feasible & valuable
• Start with a small team and expand based on the progress

AI Transformation Playbook

• Execute pilot projects to gain momentum
• Build an in-house AI team
• Provide broad AI training
• Develop an AI strategy
• Develop internal & external communications

AI pitfalls to avoid

• Don’t expect AI to solve each and everything
• Don’t depend solely on the technical team to come up with AI use-cases
• Don’t expect the AI project to work the first time
• Don’t expect traditional planning & processes to apply without any changes
• Don’t think you need superstar AI engineer to succeed

Taking your first step in AI

• Get colleagues to learn about AI
• Start brainstorming about projects
• Hire a few ML/AI people to help
• Hire or appoint an AI leader
• Discuss with CXOs/Board about possibilities of AI transformations

Survey of major AI applications

• Computer Vision: Image classification, Object recognition, Object detection, Image segmentation, Tracking
• Natural Language Processing: Text classification, Information retrieval, Name entity recognition, Machine translation
• Speech: Speech recognition, Trigger word/wakeword detection, Speaker ID, Speech synthesis
• Robotics: Perception, Motion planning, Control
• General Machine Learning: Unstructured data (image, audio, text), Structured data

Survey of major AI techniques

• Unsupervised learning
• Transfer learning
• Reinforcement learning
• Generative Adversarial Network (GANs)
• Knowledge Graph

Week 4: AI and Society

This week’s content contains all the essentials to understand the impacts of AI on our society.

A realistic view of AI

• Too optimistic: Sentient/super-intelligent AI killer robots coming soon
• Too pessimistic: AI cannot do everything, so an AI winter is coming
• Just right: AI can’t do each and everything, but will transform industries
• Performance limitations: with unavailability of data, less or irrelevant data or randomness of patterns can limit the performance
• Model explainability: explainable models are more accepted in business & society

Discrimination / Bias

• AI model may develop gender and ethnic biases
• Problem comes from the training data which contain these biases
• Examples such as:

— Hiring tool discriminating against women

— Facial recognition working better for light-skinned individuals

— Bank loan approvals

— Toxic effect of reinforcing unhealthy stereotypes

• Combating bias:

— Technical solutions: removing bias in training data, using more inclusive data

— Transparency and/or auditing processes

— Having a diverse workforce

Adversarial attacks

• AL models are sensitive to adversarial attacks
• Which are deliberate actions to fool an AI model
• It is possible by changing the values ​​of a few pixels of an image to fool a classifier
• While visually we (humans) can not detect the changes
• Adversarial defence do exist, but incur some cost
• Endless race between forgers and authorities

Adverse uses

• DeepFakes: Synthesize video of people doing things they never did
• Undermining of democracy & privacy
• Generating fake comments

AI and developing nations

• Developing economies can ‘leapfrog’ by using technology developed by others
• Like mobile phones, mobile payments, online education
• Use applied AI in specific industries
• Public-private partnerships to accelerate development
• Invent in AI education of citizen

AI and jobs

• Automation of activities by AI already has an impact on employment
• But job created by AI should be much higher than job displaced by 2030
• We should assess what are the tasks in our day-to-day job which can be automated
• Possibility of conditional basic income
• Building a lifelong learning society
• Political solutions to support AI transformation in society
• We should all complement our current knowledge with an AI knowledge

Conclusion

Let’s recap what we have learnt in this course:

• First week: AI technology, what is AI and what is machine learning? What’s supervised learning, that is learning inputs, outputs, or A to B mappings. As well as what is data science, and how data feeds into all of these technologies? What AI can and cannot do?
• Second week: What it feels like to build an AI project? What is the workflow of machine learning projects, of collecting data, building a system and deploying it, as well as the workflow of data science projects? How to carry out technical diligence to make sure a project is feasible, together with business diligence to make sure that the project is valuable before you commit to taking on a specific AI project?
• Third week: How such AI projects could fit in the context of our company? Examples of complex AI products, such as a smart speaker, a self-driving car. What are the roles and responsibilities of large AI teams? The AI transmission playbook, what are the five-steps for helping a company become a great AI company?
• Last week: AI and Society. What are the limitations of AI beyond just technical ones? How AI is affecting developing economies and jobs worldwide?

We have learned a lot in these four weeks, but AI is a complex topic, so the key is to keep learning, keep evolving.

References

AI For Everyone

Ankit Rathi is an AI architect, published author & well-known speaker. His interest lies primarily in building end-to-end AI applications/products following best practices of Data Engineering and Architecture.

Why don’t you connect with Ankit on YouTube, Twitter, LinkedIn or Instagram?

If you have any questions or comments, click the "Go To Discussion" button below!