Not just any Machine Learning framework!

Back in my childhood when I learnt swimming, nobody used fancy life jackets that kids wear today. Like Batman, I was told to face my fear and got pushed into the water, with my back first. It was a bit scary but a very effective way to learn. I totally enjoy swimming now, especially the back strokes. Infact, an age old Bateke wisdom ratifies this way of learning:

You learn how to cut down the trees by cutting them down.

In the present day, when I see organizations getting overwhelmed with Machine Learning (and other new generation techs), mostly due to ubiquitous content alluding to the prerequisites of skill set, infrastructure, data etc., to the extent that ML's value to business looks uncertain & hyped, I get reminded of my first swimming class and there is only one recommendation that sticks out.

Stop viewing ML from the edge, jump in the pool and practice.

As a student & practitioner of ML and having worked with lots of innovative companies, I'd like to share a few insights in the form of a framework that could be useful for companies to start reducing ML to practice and derive valuable outcomes.

But before I jump to the framework, I'd like to change the scene and the context a bit in the hope of providing you a better intuition for the framework.

So, tag along with me on one of my evening runs in the city of Chicago and the weekend that followed. There are some giveaways in between that you might like. If you are all set, let's go!

---

Episode 1: Serendipity (Discovery)

Summertime is winding down in Chicago, so I have started being more regular with the evening runs. One such evening, I snapped this gorgeous picture of Chicago skyline, admired it for a while, and nestled in a quiet corner to binge on some online content amidst the evening twilight.

As I scrolled down on medium.com, my eyes got fixated on the article that said – “Object Detection with 10 lines of Code”. The title piqued my curiosity to explore it fully in the weekend that was coming. I hit save and sprinted back home. (Are you still with me...?)

Episode 2: Consumption

The weekend came in quickly. I pulled up the article that essentially talked about a package called ImageAI developed by MosesOlafenwa This python library is built for empowering developers to build applications and systems with self-contained Deep Learning and Computer Vision capabilities using simple and few lines of code. The task was cut out for me. I set up the dependencies such as Tensorflow, OpenCV, Keras etc. and started reducing the code to practice using the photos from my evening runs. As I hit 'run', the output was before my eyes; accurate bounding boxes detecting objects – person, bicycle, and boat. 

Episode 3: Ideation

This activity of reducing the ready-to-use package to practice, triggered scenarios that could be useful for the companies I work with, the Industrial Machinery and Components manufacturers, to optimize or extend their business processes to address newer challenges and opportunities. I am listing some of those cases here (skip the videos in case you want to save time)

• Random bin picking tasks: Instead of manual programs, robot could learn to detect objects that need to be picked up through deep learning. Here is an example of how Fanuc, world’s largest industrial robot maker, is using deep reinforcement learning for automating tasks such as bin picking.

• Automated inventory management: Manual inventory counts is a tedious and cost intensive process replete with counting errors. Deep Learning is improving the game here. I tested ImageAI on another image I took that had lots of people. Here is what I observed.

Replace people with objects and there you have an inventory counting application. Multiple papers, example here, are being written testing applications of computer vision in this area. Below is an image from an image processing based object counting approach for Machine Vision Application; Source

• Visual inspection for quality control: Deep learning based visual inspections are turning out to be much more accurate than human inspections in cases involving product inspections not suitable to naked eyes. For example, chip manufacturing or complex assemblies built through additive manufacturing or 3D printing is ripe for computer vision applications. Landing.ai, a company started by the great Andrew NG, has developed a framework which requires very small training dataset to get started.

Many such use cases exist in other industries as well but that is a subject of another article some other day. For now, let me bring you back to the present day and present to your the framework for reducing AI to practice for valuable business outcomes. You might spot, how it could be applied to context based upskilling as well.

---

Ok! So what’s the framework to value from AI/ML?

Machine Learning/AI has reached a level where companies are genuinely exploring processes to be augmented or transformed through ML. However, not every organization has the ML/AI skill set of large technology providers and are therefore overwhelmed. In such cases, what should an organization do?

1) Foremost, understand the value chain of machine learning and your role

Know your position in the value chain of ML/AI. There are Chip makers (Nvidia, Graph core), Algorithm makers (Baidu, Google), Platform and Infrastructure providers (AWS, GCP, SCP, Azure), Enterprise solution providers (SAP, MS), Industry solution providers, and Corporate consumers (GSK, GE, Wallmart etc.). As you could imagine, your role would depend on your position in the value chain, therefore you need to approach ML/AI opportunities differently. For example, an employee in Google, which is both a producer and a consumer of AI, could either be creating newer algorithms or testing the existing ones for a new application, but an employee in a Corporate would focus more on value driven consumption by utilizing what the upstream players deliver.

As an employee, figure what is your role in the value chain of ML and set your goals accordingly.

Source for AI value chain: Best Practice AI Ltd., read more here.

2) Discover, Consume, Ideate to value.

Given the time pressure to bring new ideas to market, there is hardly any sense in getting caught in the endless loops of learning such complex technologies. Ayasdi, an AI platform company, points to the huge opportunity in the consumption of the existing AI packages and approaches. The focus need to shift to consuming the existing packages on Github, ML APIs offered by companies like MS, Google, SAP, AWS etc and to test scenarios that could bring significant benefits.

You will learn more in three days of acting than in six months of researching– Anon redditor.

The whole point of Amazon coming up with Deep Lens and Google coming up with Cloud AutoML is to let developers go crazy on finding applications of video, image, text analysis based on deep learning. The discover, consume, ideate chain that we saw with the object detection scenario above could very well be applied to your business, in turn reducing ML to practice. This could potentially be the de-facto job description of teams tasked with ML, AI tasks in such companies. 

3) Lastly, think in terms of scale and order of change

Steven Pinker, in his book ‘Enlightenment Now’ refers to scale and order of change as important criteria for any impactful policy decisions. The criteria for applications of ML, AI scenarios should be no different.

Instead of building from scratch with no scale and value in mind, you must be scratching to build for scale and value.

Post discover - consume - ideate phase, the shortlisted scenarios could be tested in terms of time & dollars spent per training or inference task versus the value that is generated for the end customer. In simple economics terms, both supply chain and consumer surplus need to be maximized. 

End note

The above framework needs to be tested more often to be set in stone but I have come across companies who do just this – they have scenarios in mind, they discover packages/papers/APIs/Github for application, they apply in their context, and look for ways to improve upon the performance only to get further inspired with more scenarios.

In other words, they take the plunge.

---

I'd love to know your thoughts and learning from practicing ML in your organizations.