CODEX
What is the difference between 30 linear steps and 30 exponential steps? In the first instance, you could probably guess that you would have travelled approximately 30 meters.
In the latter instance, you would advance at each step 1 meter, then 2 meters, 4 meters, 8 meters and so on. If you completed these 30 steps exponentially, you would ultimately reach a considerable number. In fact, you would have travelled over 1 billion meters. From another perspective — you would be able to walk around the earth 26 times and still not be done with all your steps.
These exponential growth concepts can be used to describe the extent to which Artificial Intelligence (AI) is expected to disrupt (and is disrupting) almost every industry. One main reason for this disruptive potential is its ability to overcome issues associated with scalability. There are two kinds of scalabilities: vertical scalability and horizontal scalability. We will start by considering these two dimensions from a software point of view and then expand to their potential application in the service industry.
Scalability (Software)
Scalability is the term used to describe a system’s ability to cope with increased load. However, scalability is not one-dimensional; instead, there are two ways to increase scale — vertically and horizontally.
Horizontal scalability, or scaling out, can be defined as distributing a load across multiple machines. In this case, each node contains only part of the data, and they work together to produce a greater output.
On the other hand, vertical scalability, or scaling up, can be described as upgrading to a more powerful machine. In this case, all data is stored on a single, more powerful node.
While a system that runs on a single machine is often more straightforward, high-end machines can become very expensive. Thus horizontal scalability is unavoidable. In reality, best practices usually involve a pragmatic mixture, as several fairly powerful machines can still be simpler and cheaper than a large number of small machines.
Scalability (Services)
Horizontal scalability can be seen as the distributivity of a task on a spatiotemporal level — the ability to access a service on a higher geographical and temporal plane . For example, talking face-to-face is seen as low horizontal scalability as you can only reach one person in front of you.
The next step, for example, is talking on the radio. By sending your voice over radio waves, it is possible to reach a wider audience, but this audience needs to connect at the same time to be able to listen. It means that the concept of scalability can be applied on a geographical level, but not on the temporal level.
To handle the temporal level, we can record a message, such as posting something on Facebook. You can reach a broader audience, as everyone having access to your feed can access your message at any time.
Vertical scalability can be seen as the complexity of the task — the ability to solve a complex problem. For example, if you are hurt or sick, you may try to wait it out and let your body handle the recovery.
If we increase the severity of the disease or injury, you may take some drugs to alleviate your pain or increase recovery chances.
Finally, if time and self-medication are not good enough to handle a severe disease or the issues, it would be highly recommended to see a doctor and receive their professional diagnosis.
Scalability Matrix
As we have discussed it previously, we can split scalability into two dimensions:
- Vertical Scalability, linked to the complexity of the task
- Horizontal Scalability, linked to the distributivity of the task
In the following figure, you can find a visualisation of the scalability matrix.
Following the definitions above, digital technologies, and mainly the internet, can be seen as enablers for horizontal scalability. For example, thanks to streaming services, it is now possible to listen to a recorded live concert anywhere in the world.
Conversely, AI, “Systems which perform tasks normally requiring human intelligence”, are the enablers on the vertical scalability. For example, translating languages correctly requires experts who, in addition to knowing the language, understand the message that the original text is trying to bring forward, and thus correctly transcribe it. Thanks to advances in AI, namely Natural Language Processing (NLP), it has become possible for an AI agent to perform correct translations.
Conclusion
I hope that you have enjoyed the read, in a nutshell, the concepts are relatively straightforward. There are two dimensions to scalability, horizontal & vertical, which are about distributivity & complexity, and finally, their enablers are digital technologies & AI.
The collaboration of digital technologies and AI allows this combination of horizontal and vertical scalability. Thus, by training the correct models, AI would disrupt any industry & business function with the correct data. Now, the question is, would you rather trust an AI or a human to do your bidding?
