Definition and Example:
Let’s start by defining AI. Artificial Intelligence is a field combining computer science, and mathematics giving the ability for a machine/program to detect patterns, learn from them and make decisions without being explicitly told (or coded) to do so.
For a simpler analogy, consider this: Suppose you want to create a program that predicts the weather for the upcoming days based on previous weather conditions. In normal programming, a programmer must link (code) each previous weather condition to a certain forecast, for example:
If yesterday the sky was clear & no windè next day will be clear
You can tell from above why such a forecast has a huge margin of error: The variables and combination of patterns are infinite; no human can actually code all conditions and link them to a specific forecast (no wonder you still can’t trust your local weather guy).
Enters Artificial Intelligence. With AI, a programmer can, through large amounts of data and fancy linear algebra, train a program to detect a combination of weather conditions and draw results from them. This way, our beloved weather guy can input any weather condition into our AI program which will detect the relevant pattern and give us the forecast up to 90% accuracy!
So when you hear the word AI, do not think about the Terminator telling us to get down, simply think of cool linear algebra mixed with advanced computer science enabling a program to detect patterns from large amounts of data, learn and draw results from them! Sounds simple enough, right?
If it’s that simple why is everyone talking about it? Why is it so important?
Well because this simple analogy (Ability for a machine to detect patterns and rules, learn from them and draw results) offers us infinite possibilities. You just have to look out there and see how many decisions humans make simply from previously learnt patterns and set rules:
• A doctor reads a CT scan and makes his early diagnosis based on set rules and the experience he acquired throughout the years
• A driver reacts to a certain road condition based on set rules and the experience he acquired throughout the years
• A banker assesses a loan risk margin based on set rules and the experience he acquired throughout the years
Imagine now a program that can take the experience of not one doctor/driver/banker but thousands of them, abide by set rules and draw results in a matter of seconds with a very high accuracy rate!! Well, that’s the power of AI.
With self-driving cars, AI powered scans and AI powered security systems, we are now able to eliminate fatal human errors across many industries simply because we needed a different kind of intelligence to handle these tasks for us; hint: an artificial one!
Does it already exist? If so, what’s next?
You can definitely see AI all around us. Your Netflix recommendation list, Alexa giving us the daily news, Gmail throwing spam emails in junk are all examples of AI integrated into your favorite apps seamlessly!
With the power of computers doubling every 18 months, there is no doubt that AI applications will touch every aspect of our lives within the next 20 years. In the near future, you’d be able to relax on your way home in your driverless car while an AI Chatbot is helping you choose the perfect birthday gift for your friend.
Is it dangerous? What do we have to pay attention to?
Stephen Hawking, one of the brightest minds to ever cross earth once said that AI can be the best or the worst thing that has ever happened to humankind. As mentioned before, according to Moore’s law computer power is doubling every 18 months, so it will not be long before humans can create AI programs that can outsmart themselves (An AI giving birth to a smarter AI). In these cases (what scientists call a singularity), we better make sure that we have enough regulations and rules to protect us!
But before we talk about a SKYNET scenario (If you don’t know what it is, you should watch Terminator, it’s a great movie after all), developers need to keep an eye on a lot of short-term problems which might arise from implementing AI. Especially the problem of bias: If an AI model is trained on non-diversified data, it can create racial, gender or ideological biased decisions. Imagine an AI powered airport security system that flags people from a specific ethnicity as threats just because the training data used to build the model was biased/not inclusive!
I’d like to think of AI not as artificial intelligence but rather as augmented intelligence, giving humans the power to extend their intelligence and make faster and better decisions once implemented right.
A bit of philosophy and Physics:
Max Planck, Nobel Prize physicist, was convinced that the world is deterministic, that is if at any stage we know the state of all matter in our universe, we can predict its outcome/future. Of course, quantum physics saved our free will with a probabilistic model of the universe, at least on a quantum level.
I’d like to think that with the power of AI, we can now stress test Planck’s determinism and see how much we can reduce human behavior to a combination of patterns and pre-defined rules while leaving the rest to what philosophers call free will, physicists say quantum probability and we call AI error margin.
~Rayan Najdi, COO