8 October 2019
Artificial intelligence is one of the new buzz words on the digital block.
While there is a lot of hype out there, writes Houston Muzamhindo in Analytics Consulting at Investec Private Banking, most of it has nothing to do with the apocalyptic robots we see in sci-fi movies … for now.
What exactly is AI?
There are a few misconceptions that blind us to what this technology really is. To be clear, AI is not Hollywood’s ideological idea of robots that we see in films like The Terminator and I, Robot.
Simply put, artificial intelligence is software that can interpret data, learn from data and use what it has learned from the data to achieve a specific goal.
In reality, we use AI every day. We have software that recognises both images (when you unlock your iPhone with biometrics) and recognises speech (“Hello Siri”’). When you order a meal on Uber Eats, the app uses AI to figure out the best delivery route.
In these examples, your phone is the physical casing for software which allows AI to interact with the real world, but the AI itself is simply a set of algorithms.
There are three main types of AI
① Artificial narrow
intelligence. All AI we currently use falls into this group. It is intelligence
that is built with one goal in mind and is incapable of performing any other
task – like the AI algorithm that reads your face when you
unlock your iPhone.
As sophisticated as it seems, even a self-driving car is seen as narrow or ‘weak’ AI. The AI attorney called ROSS, based on IBM Watson technology, is another example of expert but limited AI – yet one that is a close ‘mimic’ of the human brain process.
② Artificial general intelligence. We haven’t cracked this yet, but this AI will be able to have general intelligence. For example, it would be able to tell you the best lottery numbers to play while also driving your car. In other words, it could perform multiple and varied tasks that require it to have general knowledge of how things work.
In fact, it would be able to do any task a human is capable of. And who better to learn human behaviour from than a baby? Researchers are looking at how human babies learn to walk, talk, eat, jump, run and cry. If we can understand how all these tasks are learned over time, we can possibly create algorithms that can then adapt and learn new tasks when exposed to new experiences and data.
③ Artificial sentient intelligence. In this scenario, AI is not only able to perform multiple tasks but is aware of itself and will behave in a conscious and self-aware manner.
In the 2014 movie I, Robot, the robot Sonny can do every task that humans require of him, but he also becomes aware that he is “alive” and soon takes steps to preserve himself from being “killed” by his human creators. Don’t worry, it’s still the stuff of movies … for now!
So, why do we even need AI?
algorithms are applied well, AI can be more efficient and more accurate than
humans. AI is helping us do so many tasks that would have been difficult or
impossible without it. Just consider Google search results, a news feed on
Facebook, Instagram image filters and traffic control systems.
I don’t think any industry can survive without it. The key is knowing how and when to leverage it. In a decade from now, we are going to look back and wonder how our industries ever performed without AI.
Houston Muzamhindo works in Investec Private Banking in the field of Analytics and Data Science. He holds a BSc in Computational and Applied Mathematics, Wits University, PDM, Wits Business School and PGD in Machine Learning and Artificial Intelligence, Columbia University.
He is the founder of a free online education platform, IQmates, which has over 16 000 registered students. He was inducted as a TEDx Johannesburg 1830 Fellow. His favourite movie is 2001: A Space Odyssey. Follow him on LinkedIn.