How I would explain AI to my neighbor
Amazon’s Alexa; your recommended videos on YouTube and Netflix; the ads you’re seeing on Facebook; the highly viral ChatGPT. Every day we encounter applications which use artificially intelligent components. AI is becoming more endemic to practically every business sector as well as vital parts of our everyday lives. Yet what exactly is AI? Why should we care?
Being asked these questions by anyone who isn’t very familiar with AI (e.g., at a dinner table, by my neighbor, etc.) has often stumped me — it’s hard to articulate exactly what AI is without going into too much depth or without leaving so much worthy content out. Here is my attempt on paper.
What is AI?
Artificial Intelligence (AI) refers to any machine that is ‘intelligent’, i.e. solves a problem. This means that the term ‘AI’ covers everything from a rules-based chat bot (e.g., a chat bot which only answers what time of day it is given a very simple prompt), to more complicated solutions that are more likely evoked in the mind when we think about AI — robots that teach themselves to walk, chatbots that respond intelligibly to broad requests, ChatGPT, and the cool AI artworks from Lensa that people have been posting on Instagram.
AI is broader than other technical terms I’m sometimes asked about which are parts of the AI field (e.g., machine learning, deep learning, computer vision). (* see below for a brief explanation of these terms)
Where are we currently encountering AI in everyday life?
AI is included in almost every part of your digital life. Some examples:
- Apple’s Siri uses AI to listen to speech and detect what you would like to ask using a set of principles, connecting to appropriate APIs.
- Netflix, YouTube, Amazon and Spotify sift through their thousands (if not billions) of potential search results by trying to figure out what results will be most meaningful to you. For instance, YouTube is more likely to recommend you a video that has gone viral than one which has not; and if YouTube has seen you searching for a bunch of videos about cooking, you’re likely to be recommended more of those. These AI systems which sift through large amounts of data to give you recommendations are called recommender systems.
- Grammarly and other text suggestion tools are built using the machine processing of text and language (called natural language processing).
- ChatGPT and other smart chat bots are also built using the machine processing of text and language, and have processed and ‘taken note of’ extremely large amounts of data (uses what’s called a large language models).
- Ads are getting similarly prioritized in your Google searches, by AI recommender systems.
- When you play chess against your computer, you’re typically competing against an AI bot. Minecraft automatically generates data to extend its virtual worlds — this can also be considered AI.
- Credit scores are often produced using AI models based on a mix of rules and learning from other people with similar demographics to you.
Despite all these examples falling within the domain of AI, the way you design many of these systems is very different — requiring different rules and designs (called heuristics) as well as different pieces of pre-existing data to train or guide the models.
Why should I care about AI?
1. AI is becoming more and more ubiquitous in our lives, and as a great tool to turn data into insights, personalization and automation, it’s not going away. In fact, AI is expected to be a close to a half trillion dollar industry by some point between 2025–2030.
2. AI impacts decisions that we make both subconsciously and consciously. Having a basic understanding of how some AI systems are built helps explain why TikTok and Facebook are so addictive, why credit scores might contain biases as well as more positive outcomes, like why our weather predictions today tend to be so much better than in the past and why we can train a computer to recognize melanomas (a type of skin cancer) with very high accuracy. I am a firm believer that being curious about technology is the first step to building a healthy relationship with it, being able to stand up for oneself and have agency in general. You don’t necessarily have to know how to code to engage with AI, but understanding how it impacts our lives is vital.
3. AI also stands to bring unintended and interesting challenges to the forefront. For instance, consider Apple’s recent car crash detection feature: it uses AI to detect when you might be in a car crash and automatically call 911. It’s also getting set off by rollercoasters, providing unintended and unnecessary work for emergency services. Or consider autonomous vehicle services which count as one of their challenges the ‘vomit problem’, where the car is able to drive independently but human teams are required to deal with edge case logistics like cleaning up after sick passengers. The goal of these technologies are powerful and game changing, but nothing is without external implications or implementation considerations.
4. AI is raising questions about what the most important skills to learn for the future are. Just like I would argue that it’s no longer necessary to learn how to drive manual, the introduction of AI tools is changing what we need to learn about. Do we need to understand maps as well as we did in the past due to navigation tools like Google and Apple maps? Do we need to have perfect grammar and writing skills given we have Chat GPT? While navigation and communication are vital skills in general, the level to which we prioritize different skills will continue to change as AI technologies do. For instance, one likely skill to be competitive in many industries in the future is how to use no code AI tools effectively in order to maximize utility and productivity.
What other questions are on your mind? What would you add/remove from the above answers? I’d be curious to hear your take. I’ll be writing deeper dives on various topics in this article as well as some more technical articles, so let me know what might be most interesting to read.
(*) Aside — for slightly more detail
Machine learning is a type of AI that involves a machine improving on a task after exposure to more data. (Not all AI requires a lot of exposure to data)
Deep learning is a type of machine learning which learns information by progressively extracting insights at multiple layers of a model.
Computer vision is a type of AI that involves learning to recognize visual objects and features (i.e., detect information from any sort of visual medium like photographs, pictures or video). Computer vision strategies include both machine learning approaches (i.e., exposing a machine to data to learn something) and non-machine learning strategies (i.e. rule-based systems).
As a result, I like to think of the AI knowledge space very broadly as something like the drawing above, with many other subcategories not present for clarity.
All opinions expressed are my own and not of my employer or external affiliations.