There used to be a distinct, technical separation between terms such as AI and machine learning (ML) – but only while these technologies remained largely theoretical. As soon as they became practical in the real world, and then commodifiable into products, the marketers stepped in.

Widespread overuse of the terms AI/ML in marketing have managed to thoroughly confuse the meanings of these words. You might think of this as a relatively minor issue – until you realize that it’s been at the core of some thoroughly deceptive practices. Research by The Verge has shown that up to 40 percent of European startups claiming to use AI are actually lying or exaggerating their capabilities.

In short, if you don’t know what AI/ML are, or what the difference is between them, then you’re that much more likely to be sold a bill of goods when you’re shopping for a product based on those technologies. 

What is Artificial Intelligence?

There’s an automatic association between AI and sci-fi. When people think of artificial intelligence, they tend to think of the Terminator, Data from Star Trek, HAL from 2001, etc. These represent a very specific form of AI known as Artificial General Intelligence (also known as Strong AI) – a digital form of consciousness that can match or exceed human-like performance in any number of metrics. An AGI would be equally good at solving math equations, conducting a humanlike conversation, or composing a sonnet.

Currently, there is no working example of an AGI, and the likelihood of ever creating such a system remains low. Attempts to create AGIs currently revolve around the idea of scanning and modeling the human brain, and then replicating the human brain in software. This is a sort of top-down approach – humans are the only example of working sentience, so in order to create other sentient systems, it makes sense to start from the standpoint of our brains and attempt to copy them.

If you take the bottom-up approach, you end up with what’s known as Narrow or Weak Artificial Intelligence. This is the kind of AI that you see every day – AI that excels at a single specific task. AI powers apps that help you find music to listen to, tag your friends in social media photos, etc. Behind the scenes, it may help protect you or your company from fraud, malware, or malicious activity.

This kind of narrow AI does only one thing, but it does it much faster and better than a human. Imagine scanning a million purchase orders a day to make sure that there are no forgeries – you’d quickly get bored and start to make mistakes. AI could process those orders in a relative eyeblink and catch more errors and suspicious activity than even a trained human observer ever could.

What is Machine Learning?

Machine learning and artificial intelligence are not the same thing – BUT, if you’re looking to create a narrow AI the easy way, machine learning is increasingly the only game in town.

Machine learning works by getting it wrong – and then eventually getting it right. Here’s a layman’s explanation of how it works.

Let’s say you’re creating an image-recognition program in order to find pictures of cute dogs. First, you give the software program some idea of what a dog looks like. Then you show it a dataset of images – some with dogs, some without. You tell your software to pick out the dogs. In all likelihood, the software will get it mostly wrong. That’s okay. You tell the software which pictures it got right, and then repeat with different datasets until the software starts picking out dogs with confidence.

This example demonstrates a central tenet of the machine learning advantage: at no point do you have to get into the weeds of a software program and code it to recognize dogs. Instead, the machine “codes itself”, generating mathematical models to find dogs and then refining them as they’re trained on additional data.

That is the basic gist of how it works. 

When you use machine learning, you save time and effort on creating narrow artificial intelligence. Instead of creating a complex and branching decision tree by hand, your decision tree grows on its own and improves its usefulness every time it encounters and categorizes new data. By taking the grunt work out of creating models and categorizing data, machine learning vastly increases the effectiveness of data scientists.

AI/ML for Better Performance 

The difference between machine learning and AI is that machine learning represents one of – but not the only – precursors to creating a narrow AI. Specifically, machine learning is the best and fastest way to create a narrow AI model for the purpose of categorizing data, detecting fraud, recognizing images, or making predictions about the future (among other things).

Although hyperbolic marketing has in many ways distorted the meaning behind machine learning and AI, the advantage of the commodifying technology is that it’s now easier than ever to use and create machine learning models – assuming you’re working with a company that’s selling the genuine article.

Topics: Machine LearningArtificial IntelligenceAI/ML
Subscribe to our newsletter and get the latest news on AI analytics

Written by Ira Cohen

Ira Cohen is chief data scientist and co-founder of Anodot, where he develops real-time multivariate anomaly detection algorithms designed to oversee millions of time series signals. He holds a PhD in machine learning from the University of Illinois at Urbana-Champaign and has more than 12 years of industry experience.

Get a Taste of Anodot
footer background