Source: Deep Learning on Medium
Machine Learning 101 for Grannies ❤
Have you ever tried to explain computer concepts to your grandmother, grandfather, or someone who’s not very tech-savvy? It’s a little tough, right? They probably didn’t grow up with the concepts that we grew up with or that the younger generations are growing up with. We can connect to the world wide web in a matter of seconds and search for just about anything, right from the palm of our hands. I’m sure that’s still quite a mindblowing idea for some of our older generations, but we definitely take it for granted.
Now, try explaining to your grandmother what Artificial Intelligence or Machine Learning is. She may have a hard time understanding what is already a difficult concept, so let’s see if this post can help you out (disclaimer: this post isn’t just for sweet, curious grannies, it’s for anyone who wants to know a bit more about these famous concepts and is just starting out).
What is Machine Learning?
First, let’s start out by going through the definition of Machine Learning. Machine Learning is a branch or subfield of Artificial Intelligence (also known as AI). AI is a branch of computer science that aims to create intelligent machines that act or react like humans. Some of the behaviors that computers with artificial intelligence are designed to do include reasoning, deduction, learning, planning, problem-solving, and perception. They learn this through the help of machine learning and deep learning algorithms. Deep learning is a subfield of machine learning, and we’ll get into more of this later.
History of AI and Machine Learning
AI and machine learning didn’t just spring up yesterday. These computer science concepts have been around for quite some time. Here’s a quick timeline of the evolution of AI and machine learning throughout the 20th and 21st centuries:
- 1950 — Alan Turing (father of modern computing) proposes the Turing Test, in which a computer/machine and human interact, and if the human cannot tell whether or not the computer is a computer or a human, it passes the Turing Test.
- 1955 — the term AI is coined at a conference at Dartmouth College by computer scientist John McCarthy.
- 1964 — Eliza, the first chatbot, is invented.
- Some time passes before any more advances in AI occur.
- 1997 — World chess player Garry Kasparov is defeated by Deep Blue, an IBM computer capable of machine learning.
- 1998 — KISMET is invented at MIT by Cynthia Breazeal. It’s a machine that detects and responds to people’s emotions.
- 2011 — Apple’s Siri, an intelligent virtual assistant, is introduced via iPhone 4S.
- 2011 — Watson, an IBM computer, wins first place on Jeopardy!
- 2014 — Amazon launches Alexa, an intelligent virtual assistant similar to Apple’s Siri.
- 2017 — Google’s AlphaGo beats the Go world champion.
- 2019 — Google’s AlphaStar defeats professional Starcraft II players.
Types of Machine Learning
There two types of machine learning called supervised and unsupervised learning. In supervised learning, the computer is trained in the presence of a “supervisor” or “teacher” with labeled training data and other methods to produce data output from previous experiences. In unsupervised learning, the computer is given a set of data that it hasn’t seen before and it must collect information and learn from it, without interference from a “teacher”.
Difference between Machine Learning and Deep Learning
Basically, deep learning is the same thing as machine learning, but with distinct capabilities. For example, with machine learning, a computer still needs guidance to learn to accurately predict and use data. With deep learning, a computer can determine on its own whether it’s accurately using data. It progressively learns by itself. Deep learning models are designed to continuously analyze data with similar logic that humans analyze data to draw conclusions. This is done with a structure of algorithms called an artificial neural network (similar to a human brain). Due to this, deep learning is as close to true artificial intelligence as we can get so far.
Uses of Machine Learning in Our Everyday Lives
Nowadays, machine learning is used in many types of industries for various purposes. For example, machine learning can be used to accurately detect whether something in a picture is a human or an object, like in Facebook’s tagging feature. Machine learning can be used to predict which shows you want to watch based on your preferences on Netflix. It can be used to tell you how long it will take to get home based on the weather and driving conditions. It can be used in apps like Rappi to show you which restaurants you prefer based on your previous orders. There’s thousands of uses for machine learning all around the world.
Potential Dangers of AI/Machine Learning
I’m sure many of you have seen or heard of movies like Terminator, 2001: A Space Odyssey, Ex-Machina, and The Matrix, where AI is the bad guy and many times brings forth the end of the world. Well, we’re pretty far from true AI that is capable of overwriting its ethics code, but so far there have been a few instances where AI and machine learning were involved that were dangerous or deceptive for humans. For example, one of Uber’s self-driving cars was involved in a fatal accident where a pedestrian was killed by said car. In 2016, it was found that Cambridge Analytica used algorithms to exploit Facebook’s data-sharing practices to sway opinions on the 2016 US political elections. If you haven’t watched Netflix’s The Great Hack, I highly recommend it. Google employees recently went on a strike when they learned that Google was supplying the US Air Force with technology to classify drone imagery. It was feared that this could be a potential step to supply technology for automated drone strikes. In China, face recognition technology — which uses machine learning algorithms — is being deployed in the streets for government surveillance. And now, a technology called deepfake is being used to create fake and weirdly realistic clips and pictures of celebrities and politicians using neural networks. So, as you can see, not all of the uses of AI and machine learning are beneficial. However, in the long run, I believe this technology will be used in every single aspect of our lives. It’s just a matter of time.
Future Uses of AI & Machine Learning
Machine learning will continue to be used for improving our daily lives. It is used and will continue to be used in the medical, financial, tech, educational, and environmental industries. It will be used for accurate medical diagnosis, remote sensing, performing tasks that can be dangerous for humans, robot control, aviation, algorithmic trading, and many more applications. There are numerous companies around the world, including giants like Apple, Google, Amazon, Facebook, and more that are working to improve the AI experience.
One of my favorite uses of machine learning right now is Google’s Duplex model, which allows users to make appointments or reserve restaurant tables from their phone. However, the person making the phone call isn’t the user, it’s an AI which speaks directly to the employee. It uses Google Assistant (Google’s virtual assistant similar to Siri) which uses an AI-based, realistic-sounding voice. It evens does the whole “Uhh” and “Umm” that humans do when speaking. Google’s CEO Sindar Pinchai first introduced Google Duplex to the world in a May 2018 Google I/O Developers Conference. In the presentation, it appeared as though the employees on the other line never realized that they weren’t speaking to a human. Currently the service is live in almost all US states and New Zealand, where Google started a pilot program for the service. It’s also a little creepy, because of how realistic it sounds. Here’s a video on Youtube showing what I’m talking about, check it out:
It’s like some sort of Black Mirror episode. We can probably expect more of this type of application in our daily lives. Virtual assistants will become smarter and smarter and will continue to learn, until they learn all of our preferences, dislikes, behavioral patterns, etc. and will even begin to impersonate us, similar to how we saw in the video. There’s also a few uses of machine learning in the therapy world, where you can talk or chat to a “therapist” on the other side but it’s really an AI. Unfortunately, this is an example of AI taking over certain jobs. We can definitely expect to see more of this in the future and there will be a backlash against the tech community. All in all, I think the potential beneficial uses of AI far outweigh the potential downsides of it. We just need to hold the companies accountable for creating safe AI. More than 2,400 scientists and tech industry leaders have come together to sign a petition which intends to deter militaries and nations from using automated weapons, like automated drones that I mentioned earlier. It calls for a preemptive ban on these types of weapons. More than 150 AI-related firms signed the petition as well, promising to steer away from these kinds of contracts.
Thanks to several sources on the beautiful Internet, I was able to research many of the topics discussed in this post. Here they are: