A Collection of 10 Simple Yet Powerful Analogies to Understand Data & AI Concepts

Photo by krakenimages on Unsplash

In our rapidly evolving world characterized by exponential change and increasing complexity, analogies serve as invaluable heuristics — essentially cognitive shortcuts — enabling swift comprehension by bridging unfamiliar concepts with those already familiar to us. So i think knowing a few can be quite useful especially when you are trying to communicate a technical concept to a non-technical audience.

“If genius has any common denominator, I would propose breadth of interest and the ability to construct fruitful analogies between fields.”

— Stephen Jay Gould

Here’s TL;DR version of the list of 10 analogies I have come across while reading and/or interacting with customers, colleagues and friends. Hope it helps you to decide quickly if its worth your time or not.

1. Public cloud as energy utility: You can plug-and-play, and pay-as-you-go
2. Overfitting in machine learning
3. Data warehouse is like a bottle of filtered drinking water, ready for consumption
4. Tech team working in a restaurant, and their different roles
5. Multi-armed bandit for explaining epsilon-decreasing strategy in reinforcement learning.
6. AI Assistant as a Co-pilot
7. Data is the new oil (too clichéd i know, but deserves a mention and helps build the case for the next one)
8. Data strategy as a means to be able to derive valuable insights from swathes of data.
9. Containerization in shipping as well as tech
10. A few other names/terminologies directly taken from real-world examples

1. Public cloud as energy utility: You can plug-and-play, and pay-as-you-go, without having to worry about the production and distribution of electricity. Morgan Stanley analyst Katy Huberty and colleagues are the ones who seem to have quipped about this comparison in one of their reports.
2. Overfitting in machine learning: Imagine a child learning to identify chairs by looking at pictures. The more pictures they see (data), the better they become at recognizing (classifying) chairs, i.e, making predictions. Now if your training data has a vast majority of chairs having blue colour, colour as an additional dimension adds to the complexity and the child “learns” incorrectly that “all chairs are blue in colour”. Andrew Ng, in his famous machine learning course on coursera, came up with this analogy. - I think this can also be a handy analogy to explain bias in data.
3. Data Lake is like a giant reservoir holding all the natural water from various sources. On the other hand, a data warehouse is like a bottle of filtered drinking water, ready for consumption. This analogy is attributed to James Dixon, CTO of Pentaho.
4. Tech team working in a restaurant: Waiters (BAs/PMs) take user requirements and translate them into actionable tasks for the back-end. They also seek help from front-end teams who handle user interface (UI) development, ensuring a smooth and intuitive user experience. Backend developers are the kitchen staff, actually working on the dishes. Restaurant Manager (Project Manager) plans and oversees the development process, coordinating between front-end and back-end teams and ensuring deadlines are met.
5. Multi-armed bandit for explaining epsilon-decreasing strategy in reinforcement learning. Slot machines have been nicknamed one arm bandits (since they “rob” your money — house always wins). So the aim is to choose which slot machine to play in order to increase the overall payout. There’s a great book which explains this quite well.
6. AI Assistant as a Co-pilot: Imagine an AI assistant as a co-pilot on a plane, offering information and recommendations to the pilot (human) who ultimately makes the final decisions. It seems Microsoft has come up with this analogy first (not sure — please correct me if i am wrong).
7. Data is the new oil. This one is so popular that it has almost become a cliché now. British mathematician Clive Humby famously said it in 2006. He meant that data, like oil, isn’t useful in its raw state. It needs to be refined, processed and turned into something useful; its value lies in its potential.
8. Data strategy: Imagine your data is like a vast, unexplored jungle. It holds immense potential for discoveries and valuable resources, but navigating it blindly is inefficient (and fraught with risks!). So without a data strategy, you’re like - A wanderer with no map: You stumble around, collecting random pieces of information but lacking a clear understanding of the whole picture. - An adventurer with no tools: You find something valuable in it, but you may lack the ability to extract/leverage it meaningfully. By having a clear data strategy, you can unlock the true power of your data as an asset. To continue with the above analogy, having a clear and robust data strategy gives you direction, and knowing where to find specific information and resources. It also gives you the means to be able to derive valuable insights from swathes of data.
9. Containerization transformed ocean transportation by standardizing cargo units, enhancing efficiency, and streamlining loading and unloading processes. Similarly, a container in the tech world consists of an entire runtime environment: an application, plus all its dependencies, libraries and other binaries, and configuration files needed to run it, bundled into one package. Easy to move and orchestrate, just like its IRL counterpart.
10. Some other names/terminologies directly taken from real-world examples: - Dashboard, like the one found in a car, provides “drivers” with all the information they need in one place to “drive” the car effectively. - Data marts are like specialized sections within the grocery store, each catering to a specific customer group or department. A data warehouse can feed multiple data marts (as per Bill Inmon’s school of thought) - A ranger is someone who takes care of the security of a park, forest, or other public land. Similarly, Apache Ranger is a security framework for Hadoop.  - Random forest as an ensemble of multiple decision trees.

These are just a few and i am sure you would have come across plenty of more interesting ones. Feel free to share and i’d be more than happy to add them to the list (with due credits).

Cheers!