AI for Product Managers

This article is a high-level overview of AI technologies dedicated to product managers who are thinking about applying AI technology to their products.

AI Product Managers should be helpful in making decisions on what to build and whether it is feasible and valuable for the product.

All information I’ve accumulated is from AI for everyone course where lector is Andrew Ng (CEO/Founder Landing AI; Co-founder, Coursera; Adjunct Professor, Stanford University; formerly Chief Scientist, Baidu and founding lead of Google Brain).

What is AI?

According to a study by McKinsey Global Institute, AI (artificial intelligence) is estimated to create an additional 13 trillion US dollars of value annually by the year 2030.

There is a lot of excitement but also a lot of unnecessary hype about AI. One of the reasons for this is because AI is actually two separate ideas. AI technology can be divided into 2 fields:

• ANI (Artificial Narrow Intelligence) that we can see and use nowadays such as self-driving cars, smart speakers, web search, etc.
• AGI (Artificial General Intelligence) — the same actions that human can do. And it may be decades or hundreds of years or even thousands of years away.

Nowadays ANI is widely used for different projects.

AI terminology

Supervised learning.

This type of machine learning that learns A to B, or input to output mappings.

Example:

Let’s have as an input A email and we need to find out whether it is spam or not. We can use supervised learning which will filter our emails based on the given criteria.

Unsupervised learning

The reason this is called unsupervised learning is the following. Whereas supervised learning algorithms run an A to B mapping, and you have to tell the algorithm what is the output B that you want, an unsupervised learning algorithm doesn’t tell the AI system exactly what it wants. Instead, it gives the AI system a bunch of data such as, for example, customer data, and it tells the AI to find something interesting and meaningful in the data.

• Clustering algorithm is commonly used for market segmentation. It automatically groups the data and into two or more clusters.

Transfer Learning

Transfer learning is the technology that lets you from a task A, such as car detection, and use the knowledge to help you on a different task B, such as golf cart detection. Where transfer learning really shine is if having learn from a very large dataset of car detection, task A, you can now do pretty well on golf cart detection, even though you have a much smaller golf cart dataset.

Because some of the knowledge it has learned from the first task, of what the vehicles look like, what the wheels look like, how the vehicles move. Maybe that will be useful also for golf cart detection.

Reinforcemant Learning

Reinforcement learning as similar to how you might train a pet dog to behave. So, how do you train a dog to behave itself?

Well, we let the dog do whatever it wanted to do, and then whenever it behaved well we would praise it. You go, good dog, and whenever it does something bad you would go, bad dog. And overtime it learns to do more of the, good dog, things, and fear of the bad dog things.

Reinforcement learning takes the same principle and applies it to a helicopter or two other things. So, we would have the helicopter flying around in a simulator so it could crash without hurting anyone.

GANs

GANs (Generative Adversarial Network) are very good at synthesizing new images from scratch. Everything ranging from computer graphics to computer games, to media, and to just making up new contents like this from scratch.

You can see the example here.

Knowledge graph

If you do a search on Google of Leonardo da Vinci, you might find this set of results with this panel on the right of information about da Vinci.

This information is drawn from a Knowledge Graph, which means a database that lists people and key information about these people. Such as their birthday, when they pass away, their bio, and other properties of these individuals.

The term Knowledge Graph was initially popularized by Google, but this concept has spread to many other companies. Interestingly, even though Knowledge Graphs are creating a lot of economic value for multiple large companies at this point, this is one subject that is relatively little-studied in academia.

NLP

NLP (Natural Language Processing) is used in different areas, such as:

• Text classification (email -> spam/non-spam, product description -> product category)
• Information retrieval (web search)
• Name entity recognition
• Machine translation
• Others (parsing, part-of-speech tagging)

Data Science

Data science can extract knowledge, new insights from the data. This is the main difference from machine learning, where input and output model is used.

Data science can analyze data and suggest hypothesis or actions.

Example:

If analyzing data tells you, for example, that the travel industry is not buying a lot of ads, but if you send more salespeople to sell ads to travel companies. You could convince them to use more advertising and the executives deciding to ask a sales team to spend more time reaching out to the travel industry.

Deep learning (Neural Network)

It is a big mathematical equation. No, it does not work like human’s brain. Though, we don’t even know exactly how our brain works. This type of AI can analyze big amounts of data calculating it and extracting the answer.

Example:

You need to predict the demand of the T-shirts. You collect the data and analyze all the dependecies. Based on this information you can predict whether your T-shirts will be popular or not.

What AI can and cannot do

Machine learning tends to work well when:

• Learning a “simple” concept
• There is lots of data available

Machine learning tends to work poorly when:

• Learning complex concept from small amount of data
• It is asked to perform on new types of data

Workflow of AI projects

Here are the examples of how machine learning and data science projects work. You can see their difference, how they work and what is the result.

Key steps to work on Machine Learning project:

1. Collect data
2. Train model (iterate many times until good enough)
3. Deploy model

• get data back
• maintain/update model

Key steps for Data Science project:

1. Collect data
2. Analyze data (iterate many times)
3. Suggest hypothesis/actions

• deploy changes
• re-analyze new data periodically

How to choose AI?

Before choosing AI think of the necessity to apply it.

Brainstorming framework:

1. Think about automating tasks, rather then automating jobs. E.g., call center routing, radiologists, etc.
2. What are the main drivers of business value?
3. What are the main pain points in your business?

Example roles of an AI team

AI team can include such roles:

• Software Engineer
• Machine Learning Engineer (A -> B)
• Machine Learning Researcher (extent state-of-the-art in ML)
• Data Scientist — this role is not determined exactly. In general it is supposed that data scientist should examine data and provide insights; make presentation to the team and executives
• Data Engineer — organize data; make sure data is saved in an easy accessible, secure and cost effective way.
• AI Product Manager — help decide what to build; what’s feasible and valuable.

AI technical tools

Machine learning frameworks

• TensorFlow
• PyTorch
• Keras
• MXNet
• CNTK
• Caffe
• Paddle Paddle
• Scikit-learn
• R
• Weka

Research publications

• Arxiv

Open-source repositories

• GitHub

Steps to add AI to your project

1. Execute pilot.

• More important for the initial project to succeed rather than be the most valuable
• Show traction within 6–12 months
• Can be in-house or outsourced

2. Build an in-house AI team.

New AI business unit can be under the CTO, the CIO, the chief data officer or the Chief Digital Officer or it could also be under a new chief AI officer.

3. Provide broad AI training.

4. Develop an AI strategy

• Consider creating a data strategy (strategic data acquisition, unified data warehouse)
• Create network effects and platform advantages (in industries “winner take all” dynamics, AI can be an accelerator)
• Leverage AI to create an advantage specific to industry sector
• Design startegy aligned with the “Virtual cycle of AI”

5. Develop internal and external communications.

• Investor releations
• Government relations
• Consumer/user education
• Talent/recruitment
• Internal communications

AI pitfalls to avoid

Don’t:

1. Expect AI to solve everything.
2. Hire 2–3 ML engineers and count solely on them to come up with use cases.
3. Expect AI project to work the first time.
4. Expect traditional planning processes to apply without changes.
5. Think you need super-star engineers before you can do anything.

Do:

1. Be realistic about what AI can and cannot do given limitations of technology, data and engineering resources.
2. Pair engineering talent with business talent and work cross-functionally to find feasible and valuable projects.
3. Plan for AI development to be an iterative process, with multiple attempts needed to succeed.
4. Work with AI team to establish timeline estimates, milestones, KPIs, etc.

Conclusion

AI is a powerful tool, but not all the products really need it. Think carefully whether you need it or not.

Hope it was useful for you! Would be glad to discuss it with you here or on my LinkedIn profile https://www.linkedin.com/in/epavlovskaya/.