Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

Ask Not What You Can Do for Machine Learning...

DZone's Guide to

Ask Not What You Can Do for Machine Learning...

My fellow product managers, the time is upon us to ask not what you can do for machine learning; ask what machine learning can do for your product.

· AI Zone ·
Free Resource

There is a great read by Ed Newton-Rex that outlines the differences between artificial intelligence and machine learning in simple terms. It might be good to read before you read the rest of this post!

This recent article from Frank Chen on the expected use of artificial intelligence in startups highlights an upcoming shift in the industry. He compares, and rightfully so, the pervasiveness of AI to other game changers in the industry such as having a relational database in 1980 or rich Windows client in 1987 and so on.

AI is nothing new in computer science — modern AI research was born in 1956 at a Dartmouth workshop (coincidentally included two, now-famous, Carnegie Mellon professors with a building named after them — Allen Newell and Herbert Simon). Even though we're still some ways away from building our one true AI overlord (AKA Skynet), over the past two decades, we have seen significant advances in the field of machine learning (ML). However, the practical applications of this technology so far have been elusive with only select few large enterprises having the resources and skills necessary take advantage of it.

Where do you see ML being applied in your product? If you have started using these capabilities, what has your experience been like? I would love to hear stories and thoughts in the comments below. Special thanks to Jason Cahill and Neel Kishan for their input.

  • Is there a large volume of data (structured or unstructured) involved in any part of your product lifecycle? In 2017, the likely answer would be yes. However, I have come across a few rare cases where despite the existence of a large data corpus, the use case was too direct that incorporating ML today was an overkill, i.e. visual representation based on results of complex SQL queries.
  • Does your product involve humans as part of its workflow? This is particularly true for heavily regulated industries that require frequent auditing of intermediate artifacts, i.e. reading insurance claims before making the final decision on payment.
  • Does your product deliver the final outcome of the problem it is trying to solve? This is a tricky one and requires you to truly break away from any preconceived notion of, "What is possible?" In most cases, you will find the answer to be "yes." However, that is good news because you can use ML to get your product one step closer to customer's end goal. A good example is solving delivery for an online retail store — the traditional approach is to acquire more delivery resources with greater operational efficiency but imagine if your app not just takes your order but picks up your item from the nearest warehouse and delivers it to your door (or roof) using a drone. Computers are already writing news articles, winning at Go, and creating music — maybe your idea isn't too farfetched.

That is changing. Since the beginning of this decade, cheaper access to processing power on the cloud has finally brought about the tipping point where you no longer need ML specialists to make use of these advanced ML capabilities — whether it is identifying faces in your photo app or understanding the context of a question asked in natural language. These "off-the-shelf" services that use ML can help you design an exciting experience for your customers at a fraction of the cost.

More importantly, though, the inclusion of machine learning fundamentally changes the traditional approach to product management and design. Products can now be made to achieve greater automation for tasks that have required human intervention in the past. The task of knowing whether a photo has a mountain or a skyscraper no longer requires a human tagging the photo neither does the task of identifying relations and named entities in unstructured text.

This is having a profound impact on how the next generation of software products are defined and, as a result, why the next generation of software product managers need to think about what ML can bring to the table. As a manager of products that rely on machine learning, I regularly come across use cases from a variety of industries such as healthcare, insurance, telecommunication, and intelligence. Of course, the applicability of the technology varies industry by industry but there are few key aspects that I have found helpful as a product manager to look out for when identifying where ML can be used. The list below is, by no means, exhaustive.

If you answered "yes" to any of the questions above, there is a chance machine learning might be able to help your product. No, I am not advocating staying late tonight and rewrite your current feature roadmap to include ML everywhere. I am simply suggesting an acknowledgment of these new tools at your disposal and think about how they can augment your product in future.

Each product is different and there won't be a solution that fits all but my hope is that answering these questions would get you started thinking about ways you can incorporate this technology as part of your product design.

In summary, my fellow product managers, the time is upon us to ask not what you can do for machine learning; ask what machine learning can do for your product.

Topics:
ai ,machine learning ,product management

Published at DZone with permission of

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}