Artificial Intelligence — Asking for a friend.
For the last decade, we’ve found ourselves using words like “Artificial Intelligence” and “Machine Learning” as something that either corresponds with a futuristic, sci-fi way of life with self-driving cars, holograms, and super complex software that only tech giants have the capacity to build and access - or as something more aligned with a robot uprising and destruction of the human race 🤖.
The truth is, a lot of us don’t really know what Artificial Intelligence is.
“Genius is making complex ideas simple, not simple ideas complex.”
— Albert Einstein
Having spent the last 8+ years in tech, I’m no stranger to fancy jargon and enough acronyms to MYHS (make your head spin). Have I been a culprit of overusing them myself? Guilty — you live and you learn! That said, my most productive, forward-thinking, and fulfilling work has been the result of simplifying complex concepts, taking the time to understand the foundations and basics, and deeply thinking about use cases so I can build solutions that matter, that have the ability to scale, and where I can make real, tangible progress with my work.
While I’m hardly an expert on the matter of Artificial Intelligence, I figured there are others out there nervously smiling in conversations with buzzwords that aren’t totally making sense (👋), or those looking to get involved in AI but don’t know where to start.
Wherever you fall on that spectrum, let’s work through this 101 together.
What is Artificial Intelligence? Asking for a friend.
Artificial Intelligence is really just a software program that aims to mimic human intelligence. On a high level, AI software programs have been created based on a deep study of how the human brain works, learns, thinks, reasons, decides, and functions to perform human-like tasks — at significantly larger scales and faster speeds.
One of the best parts about AI is its ability to free up humans by taking over and automating monotonous and time-consuming tasks that would normally be performed by a human. This creates the time and space for humans to further innovate, and has had real-world implications at improving the lives of many groups of people.
Got it. So what does AI actually do?
To simplify, AI is essentially good at three main tasks :
- Ingesting and processing copious quantities of data
- Recognizing patterns within that data
- Making predictions (or decisions) based on that data
What you might have noticed above is the codependency of data and AI. Data (big or small) is dependent on AI’s ability to process, recognize patterns, and build out predictions and models to be effective and actionable. AI is dependent on data to exist.
It’s important to call out here that while AI can work with massive amounts of data (often referred to as “Big Data”), it can also work with smaller amounts of data. This should have you thinking about applications of Artificial Intelligence for various business functions, big or small.
My next hot take : Garbage in, garbage out. The key here is the quality of data — and the next big real problem for us to solve is the method of data collection, deliberate or otherwise. More on that later, so keep reading for now.
How does AI use data to make predictions, decisions, reason, etc.?
There are several techniques that go into building and enhancing Artificial Intelligence programs (sometimes considered subsets of AI), with the aim to replicate human brain functions. A few examples include :
- Machine Learning (another buzzword, often also referred to as “ML”): This is the science of training AI systems to use large volumes of data and algorithms to draw inferences, classifications, predictions, and decisions from patterns found in that data.
- Deep Learning (an arm of Machine Learning) : This is the technique that uses statistical modeling to teach and train the Machine Learning algorithm to self-learn, and adapt and adjust algorithms without requiring human intervention. Think : Facial recognition on social media.
- Neural Networks (a subset of Deep Learning) : Simply put, this is a vast network of machines that resemble and work like the web of neurons in the brain.
- Fuzzy Logic Systems (used with Neural Networks) : This is an approach to “degrees of truth”, rather than the binary true / false, and resembles human reasoning to make decisions. Think : Self-driving cars can use fuzzy logic to determine the distance (not just close vs far, but rather a scale of close to far) from the car in front of it to decide what pressure to apply on the breaks. Visual reference for those like me👇
Keep in mind, the architecture of these techniques is a lot more nuanced and sophisticated than how I’ve described them here, but this should be a good starting point.
A few Artificial Intelligence tools and frameworks you may hear of in conversation include : TensorFlow, PyTorch, Auto ML, MATLAB, Google ML Kit, and more.
I need another example to understand this. Preferably one that has to do with 🐶
The Furbo pet-monitoring camera’s Furbo Dog Nanny (FDN) service (or really, any other smart camera service) is one of many interesting examples of an application that is improved using AI and deep learning. Let’s dig in. According to the Furbo website :
Furbo Dog Nanny is a premium dog-monitoring service that helps you look after your dog and keep them safe at home. The service includes Smart Dog Alerts, Cloud Recording, and Doggie Diary.
The Furbo app asks you to help train the “Dog Nanny” to recognize patterns in its video recordings. It’ll show you the view from your camera, and ask you to confirm if the object detected is a “person” or what your dog is doing — “walking”, “choking”, “chewing” and so on. This continuous data feedback teaches the Furbo application through deep learning techniques to detect and recognize those objects and actions on its own so it can create appropriate security protocols and send a notification to your phone accordingly — “Your dog has been barking for one continuous minute” or “A person has been detected”.
This makes it easy for anxious pet owners to leave their pets at home, knowing their smart AI camera will send them a notification if and when there is any concerning activity detected.
Key takeaway : Humans now have that time freed up for other tasks.
Nice. Why are we all so excited about AI?
The potential benefits of AI are limitless at taking over repetitive mundane tasks, making lives more efficient, speeding up processes, and freeing up time for more innovative and important matters. The potential expands across all industries, including farming and agriculture, finance and investments, energy efficiency, retail inventory management, education, environmental protection, resource management, space exploration, waste management and more.
We’re already using AI across so many different applications and systems in our daily lives, we might not even notice a lot of them. Do you use an automated investment or trading tool? Do you enjoy getting recommendations for new music you might like? Does the facial recognition feature on social media save you time searching for names to tag? How about chatbots that help you find the information you need on a website at any time of the day? Google giving you search results despite your terrible typos? Yeah, that’s AI.
This all sounds great! Why are we still concerned about AI?
This is a much deeper conversation around ethics in technology. There are questions about how much diversity and representation there is in the techniques for building and enhancing Artificial Intelligence, the prejudice (intentional or not) in data gathering methods, and what level of unconscious bias may be applied to machine learning. These factors will ultimately determine the “output” — the reasoning, predictions, and decisions made — the direction of our society.
Let’s consider the example of AI-powered content recommendation based on identity graphs created and designed by a non-diverse group of individuals. These identity graphs might stereotype people based on race, religion, gender, nationality, and so on — and only show them content that fits those stereotypes. It’s likely these content recommendations can further divide and isolate societies by fueling discrimination and perpetuating harmful narratives, but also limiting growth and exposure to content that just wasn’t a fit for your “classification”. The same concept of algorithmic bias can be applied to countless other areas where AI is being used to make predictions and decisions, including the criminal justice system, organizational hiring, pharmaceuticals and healthcare, and more.
There is also the philosophical challenge of human judgement. Any error in decision-making or prediction in healthcare based on AI can have a devastating impact on the patient, which is why human involvement between doctor and patient remains so important while working with digital technology leveraging AI.
A similar concern and concept applies to creative process and art. While recent news has been exploring the potential threat of AI to artists and their work (recently : Taylor Swift, Joe Rogan & Steve Jobs, and also image-generating platforms like Dall.e), my viewpoint here is that art will always require an element of raw human creativity, expression, and connection so I think (hope!) we might be good on this front. For now, at least.
Unemployment is another concern you’ll often hear of, as AI starts taking over human jobs. While it is true, a positive outlook on this front is that each time technology has taken over tasks or jobs from humans, it has created time and space for us to think and do more and create new kinds of jobs. One could argue the introduction of washing machines and refrigerators into family homes decades ago served as the catalyst to bring women into the workforce.
“If I had asked people what they wanted, they would have said faster horses.”
- Henry Ford
While there are several other concerns around AI, the simulation of human judgement and reasoning in robotics is another area demanding our focus (looking at you LaMDA 👀, and you Westworld). As we continue down the path of creating robots, there is a stronger need for ethical standards that may have to be encoded into AI agents, with a consolidated effort to remove as much bias as possible.
This feels like an oversimplification of Artificial Intelligence.
You’re right. Because it is.
Ideally, this simplified explanation of AI will serve as a thought starter, or at the very least, as a cheat sheet for a more informed conversation at your next happy hour.
If we are to work towards the democratization of AI and AI tools as a tech community, we need to simplify the complexity of it for more than just Engineers to help business functions move forward. The key to moving forward effectively is to invite diversity of thought and perspectives into the conversation.
Like any new technology, there are always benefits and drawbacks to consider, along with important elements to keep in mind as we move forward on a certain path. There are certainly parts of where AI is headed that are concerning, and some downright creepy; but with the right amount of course correction and thoughtfulness, there are also so many incredibly exciting opportunities with AI that can unlock the potential to greatly improve lives and societies across the world.
Given that we’re already so far in that it’s not even realistic to go back now, it’s important we reframe our thinking at this point in time from a fear-based perspective of : “What should we do now!?”
— to a possibility-based perspective of :
“What could we do now?”