Discovering Innovative Deep Learning Solutions

The first step in developing an AI solution is to first identify a customer’s needs. I’ve begun to use Wardley Maps in my analysis of business models that employ Deep Learning as a critical component. Wardley Maps is well suited for other business models but it is uniquely valuable in that it includes knowledge that technology exists at different stages of maturity.

The reason why I find this important, in relation to emerging technology like Deep Learning based Artificial Intelligence, is that it is rare that solutions of value are created completely from scratch. Technology evolution moves at the pace that human organizations can reorganize their own processes to best leverage the new. I recommend reading Brian Arthur’s “The Nature of Technology and How it Evolves” to gain a clearer understanding of how technology evolves and influences society.

I previously described how to identify value using the Jobs to be Done (JTBD) framework. Clay Christensen who is more known for writing “The Innovator’s Dilemma” has written a prescription on how to uncover innovative products. His prescription is a “Theory of Jobs to Be Done”. JTBD framework is method for uncovering the tacit needs of customers. In conventional approaches, we segment customers through attributes such as income, age, race, and other categories and create products based on this segmentation. The focus is misplaced in that it emphasizes want a company wants to sell rather than focusing what a customer actually needs. Christensen explains this best in his this story about milkshakes:

Instead of focusing on attributes of the milkshake (thickness, amount of syrup, etc.) the researchers sought to understand the Job to Be Done for milkshake customers. They learned that customers were “hiring” the milkshake to help them stave off hunger and avoid a mess during their morning commute. The competitors for this job weren’t other milkshakes, but breakfast foods that are easy to consume while driving, such as bagels and bananas. With this insight, the fast-food chain began marketing the milkshake as a breakfast item, and sales soared.

Customers purchase goods because they find themselves needing it to solve their specific tasks. A product is used when it is recognized as one of the tasks needed to reach a goal. Therefore, we need to understand the context in which a customer finds himself in and then identifying solutions that enable that customer to achieve their goals. Understanding the “Job to be Done” leads towards the creation of products that are more likely to be “hired”. Useful and thus valuable products are tailored to addressing what customers need to get done.

JTBD is a useful framework for both physical as well as virtual goods. Virtual goods, however, are particularly problematic in that they intrinsically lack scarcity. Virtual goods can be copied ad infinitum. But, scarcity is a necessary ingredient for all economics. How then can we re-introduce scarcity back into the value equation so that we can build a sustainable business model? How do we have customers pay for goods that aren’t scarce?

Kevin Kelly has 8 answers to this conundrum. Kevin Kelly’s book “The Inevitable” describes technology trends that are inevitable. We shall see the increase of sharing, filtering, tracking, screening, interacting, accessing, interacting, remixing, cognifying, questioning and flowing. Many of these trends are self-evident. However, Kelly has unique insight into what he describes as flowing.

He describes eight generatives that are “better than free.” The answer is to embrace the frictionless flow of virtual goods by leveraging its natural liquidity. He writes “success in this new realm requires mastering the new flow.” Kelly’s generatives identify new kinds of scarcity:

IMMEDIACY — Although you can eventually find a free copy of a virtual product, there is value of receiving it as the moment of its release. An audience values content when its delivered hot-of-the-presses.

PERSONALIZATION — A consumer may pay more for a modified version of the original content that is customized to the user’s preference. Personalization creates stickiness in that it generates a unique relationship between the producer and its users that is difficult to sever.

INTERPRETATION- Most open source business models are based on free software but paid support and education. The interpretation of a bunch of bits, what it means what you can use if for and how you can use is valuable for complex products.

AUTHENTICITY — Companies that employ virtual products of critical importance will pay for peace of mind of paying for the authentic version.

ACCESSIBILITY- There is value in having services conveniently accessed in different contexts. As an example, a book can be physical, in electronic form, in audio form and perhaps in the future in many alternative forms.

EMBODIMENT — There is a reason why people will pay an extravagant amount to be in a attend a music event in person compared to a version streamed via the internet. Humans value the experience of being embodied with the content.

PATRONAGE — Fans have an interest in being associated with their favorite content.

DISCOVERABILITY — Masterpieces that cannot be found are effectively worthless.

To understand how we can employ Deep Learning to enable each of these generatives, we leverage the perspective of identifying tasks with cognitive load. To recap, the four classes of cognitive load that have previously identified are information overload, lack of meaning, limited memory and the need to act fast. What we recognize is that each of these generatives, can be more relevant to a customer when we have a solution that reduces cognitive load.

Goods are acquired by human beings because they fulfill jobs to be done. But what is the universal job to be done that all humans are driven to fulfill? I will state that this universal job is the need for sharing experience. Research by Tomasello has shown that “shared intentionality” is a cognitive capability that is unique to humans. The great apes have advanced cognitive abilities. In some areas, they are gifted with superior capabilities than humans. Chimpanzees have larger short term memories and can react much faster than humans. However, it is our ability for shared intentionality that distinguishes humans from the great apes.

Ask yourself, what makes for compelling art? Art has an aspect where an artist shares an experience with his audience. Beauty indeed in the eye of the beholder, but it is the artist that generates the masterpiece that affords a beholder a route to empathize with an artist’s experience. All meaningful human endeavor is directed towards shared experience. One will have difficulty deriving human meaning without sharing with other humans. The need originates from our mammalian biology but it is further attenuated by our unique cognitive capability of shared intentionality.

Humans navigate our everyday lives from one shared intentional space to another successive shared intentional space. We hit the road to work, coordinating our movements with others to most efficiently get to our destination. At work, we are constantly coordinating our tasks with the tasks of other works. When we leave for work, we are again coordinating our activities with our family or our friends. When we got to bed, our minds are consolidating our experiences to better deal with the coordinations required in the future. This is why the Japanese have formulated this idea of Ikigai (生き甲斐): A reason for being:

Participating in shared intentional spaces demands commitment for its participants. The essence of Ikigai is that it involves a commitment of time and beliefto a particular cause, skill, trade, or group of people. There is no being without a commitment to sharing. Commitment implies sacrifice and it is the sacrifice that is the price of admission to a meaningful life.

So now we have a general framework that consists of (1) a reason for being that is intrinsic in most humans and (2) the have cognitive limitations that hinder our movement to getting things done. How can we use cognitive technology like Deep Learning to enhance the 8 generatives needed for virtual products? How can we increase the chances that our virtual products are hired?

IMMEDIACY- Why is immediacy important for shared intentionality? Individuals that are first to popular shared experiences are uniquely positioned to share their interpretation of that experience. Services that can curate the multitude of available experience is valuable. There is value in arriving in a place and instantly knowing what events are meaningful for your attention and what should be ignored for a later time.

PERSONALIZATION- Products that get hired need to be personalized for its hiring user. The conventional method of reducing costs is to commoditize a product. Unfortunately, although one size fits all reduces manufacturing cost, it adds cognitive friction for its users. What we need are products that can dynamically adapt itself to the cognitive processes of its users. We need cognitive tools that are empathic to the preferences of its users. Personalization enhances an individual's ability to be more efficient in achieving one’s commitments. Personalization reduces the cognitive load of knowing how to use or interact with a product. It reduces information overload.

INTERPRETATION — Complex products require cognitive tooling to bridge the knowledge gap between a user’s goal’s and how it can be achieved. Advanced NLP methods can be used to transform content into easier digestible chunks. Interpretation reduces information overload, speeds up the adoption of new tools and repurposes information into more meaningful/useful formats.

AUTHENTICITY — In a world with information overload, consumers value cognitive tooling that differentiates useful information from noise.

ACCESSIBILITY — Customers will demand access to products in different modalities. It becomes prohibitively expensive to transform content. Deep Learning has advanced the ease of creating richer derivative content. As an example, one can get high-quality speech rendition today. This addresses the cognitive load to quickly generate content.

EMBODIMENT — Humans have evolved to favor embodied interactive experiences. Learning is, in fact, more effective through the act of participation and doing. Deep Learning combined with virtual or augmented reality interaction can enhance original content. This addresses the cognitive load of facilitating the revelation of meaning.

DISCOVERABILITY — We know this technology in the form of search services that’s able to dynamically find relevant information to our queries. These are typically enabled by Deep Learning technologies.

The first step in leveraging Wardley Maps begins with identifying the needs of the customer. This is much easier said than done. What I’ve provided here is a more informed framework of how to identify these needs from the perspective of employing the latest cognitive technologies.

As a reminder, Deep Learning can be extremely valuable in contexts that integrate with the physical world (i.e. monitoring power lines, self-driving cars etc.), this article confines its scope solely on virtual goods.

Further Reading

Strategy for Disruptive Artificial Intelligence: https://deeplearningplaybook.com