The difference between AI and machine learning, explained

Some time ago, while I was browsing the latest news on AI, I came across a company that claimed to use "machine learning and advanced artificial intelligence" to collect and analyze hundreds of data touch points to improve the user experience in mobile applications.

The same day, I heard about another company that had predicted the behavior of its customers using "a combination of machine learning and artificial intelligence" and " predictive analysis based on AI.

(I will not name companies for shame, because I think their products solve real problems, even if they market them in a misleading way.)

Artificial intelligence and machine learning are very confusing. Some people refer to AI and machine learning as synonyms and use them interchangeably, while others use them as separate parallel technologies.

In many cases, people who speak and write about technology do not know the difference between AI and ML. In others, they intentionally ignore these differences to create hype and excitement for marketing and sales purposes.

As with the rest of this series in this post, I will try to understand the differences between artificial intelligence and machine learning to help you to distinguish facts from fiction. as regards AI.

We know what is machine learning

Let's start with machine learning, which is the easiest part of the AI ​​vs ML equation. Machine learning is a subset of artificial intelligence one of the many ways you can run AI.

Machine learning relies on the definition of behavioral rules by examining and comparing large sets of data to determine common patterns. This approach is particularly effective in solving classification problems.

For example, if you provide an auto-learning program that includes a large number of X-ray images and the corresponding symptoms, it will be able to assist (or possibly d & # 39; automate) the analysis of x-ray images in the future.

The machine learning application will compare all these different images and will find what are the common patterns found in labeled images with similar symptoms. And when you provide him with new images, he compares his content with the patterns gleaned and tells you how likely it is that the images contain one of the previously studied symptoms.

This type of machine learning is called "supervised learning," where an algorithm entails on human-labeled data. Unsupervised learning, another type of ML, involves giving the algorithm untagged data and allowing it to find patterns by itself.

For example, you provide an ML algorithm with a constant flow of network traffic and let it learn for itself what is the basic and normal activity of the network and what are the aberrant and possibly malicious behaviors occurring on the network.

Reinforcement learning, the third most common type of machine learning algorithm, consists in providing a ML algorithm with a set of rules and constraints and allowing it to learn by itself to achieve better its objectives.

Reinforcement learning usually involves some sort of reward, such as scoring points in a game or reducing the consumption of electricity in an installation. The ML algorithm does its best to maximize its benefits within the constraints provided. Reinforcement learning is famous in for teaching artificial intelligence algorithms to play different games such as Go, poker, StarCraft and Dota.

Machine learning is fascinating, especially its more advanced subsets such as deep learning and neural networks. But it is not magic, even if we sometimes have a problem to discern its internal functioning .

Basically, ML is the study of data to classify information or predict future trends. In fact, while many like to compare in-depth learning and neural networks to how the human brain functions, there are great differences between the two .

Result: We know what is machine learning. It's a subset of artificial intelligence. We also know what he can and can not do .

We do not know exactly what AI is

On the other hand, the term " artificial intelligence " has a very broad scope. According to Andrew Moore Dean of Computer Science at Carnegie Mellon University, "artificial intelligence is the science and engineering of making computers that, up to # In recent times, it required human intelligence.

This is one of the best ways to define AI in one sentence, but it still shows how wide and vague the field is. For example, "until recently" is something that changes over time.

Several decades ago, a pocket calculator was considered an AI because only the human brain was able to calculate it. Today, the calculator is one of the most stupid applications you will find on all computers.

As Zachary Lipton, publisher of Approximately Correct explains, "the term artificial intelligence" is an aspiration, a mobile target based on the capabilities that own the humans but the machines do not have. "

Artificial intelligence also encompasses many technologies that we know. The machine learning is only one. Earlier AI work used other methods such as the good old-fashioned IA (GOFAI), which is the same if-then rules that we use in other applications. Other methods include A *, fuzzy logic, expert systems and much more.

Deep Blue, the artificial intelligence that defeated the world chess champion in 1997, used a method called tree search algorithm to evaluate millions of hits each turn.

Many of the references made to AI relate to general AI or human-level intelligence. This is the kind of technology that we see in sci-fi movies such as Matrix or 2001: The Space Odyssey.

But we still do not know how to create an artificial intelligence comparable to the human mind, and deep learning, the most advanced type of AI, can compete with a child's mind, let alone 'an adult. It is perfect for narrow tasks and not for general abstract decisions which is not a bad thing at all .

The AI ​​as we know it today is symbolized by Siri and Alexa, by the extremely precise film recommendation systems that animate the Netflix and YouTube algorithms by the used by hedge fund algorithms. to make micro-trades that earn millions of dollars each year.

These technologies are becoming more and more important in our daily lives. In fact, it is the technologies of augmented intelligence that strengthen our capabilities and make us more productive.

Conclusion: Unlike machine learning, artificial intelligence is a moving target and its definition changes as its associated technologies become more advanced. What is not an AI can easily be challenged, which machine learning is very clear in its definition. Maybe in a few decades, advanced technologies in artificial intelligence will be considered as stupid as calculators are today.

So, if we go back to the examples mentioned at the beginning of the article, what does "machine learning and advanced artificial intelligence" really mean? After all, do not we learn in machine and depth the most advanced AI technologies currently available? And what does "predictive AI-based analysis" mean? Does predictive analytics not use machine learning, which is a branch of AI?

Why do technology companies use both AI and ML?

Publications use images such as crystal balls to give AI an aura of magic. This is not it.

Since the term "artificial intelligence" was coined, the industry has seen many ups and downs. In the early decades, there was a lot of hype around the industry and many scientists promised that human-level AI was imminent.

But the broken promises caused a general disenchantment with the industry and led to the winter winter period during which the funding and interest for the sector collapsed considerably.

Next, companies tried to disassociate themselves from the term "AI", which became synonymous with unfounded hype, and used other terms to refer to their work. For example, IBM described Deep Blue as a supercomputer and explicitly stated that did not rely on artificial intelligence contrary to what it was technically

During this period, other terms such as big data, predictive analytics and machine learning began to gain ground and gain popularity. In 2012, machine learning, deep learning and neural networks made great strides and began to be used in a growing number of areas. Companies have suddenly started using the terms machine learning and deep learning to market their products.

In-depth learning started performing tasks that were impossible to perform with rule-based programming. Areas such as speech and face recognition, image classification and natural language processing that were at a very crude stage, suddenly made big leaps .

And maybe that's why we are seeing a return to AI. For those who were accustomed to the limits of traditional software, the effects of deep learning seemed almost magical, especially since some of the areas in which neural networks and computer systems were used to learn more about it. In-depth learning enter were considered forbidden for computers.

Machine Learning and Advanced Learning Engineers Earn – 7-Digit Wages Even When They Work for Non-Profit Organizations, Evidence of the domain strength. ]

Add to that the erroneous description of neural networks, according to which the structure mimics the functioning of the human brain, and you suddenly have the feeling that we are moving again towards artificial general intelligence. Many scientists (Nick Bostrom, Elon Musk …) have begun to warn of an near apocalyptic future where ultra-intelligent computers drive humans into slavery and extinction. Fears of technological unemployment resurfaced.

All of these elements helped revive the excitement and hype surrounding artificial intelligence. Therefore, commercial services find it more profitable to use the vague term AI, which has a lot of luggage and gives off a mystical aura, instead of being more specific about the type of technologies that they use. . This helps them sell or resell the features of their products without realizing their limitations.

Meanwhile, the "advanced artificial intelligence" that these companies claim to use is usually a variant of machine learning or another known technology.

Unfortunately, it is common for technical publications to report these facts without careful analysis, and they often accompany Amnesty International articles with images of crystal balls and other magical representations.

This will help these companies generate hype around their offers. But in the future, not meeting expectations, they are forced to hire humans to fill the gaps of their AI . In the end, they could end up causing mistrust on the ground and trigger a new AI winter for short-lived gains.

This story is republished in TechTalks the blog that explores the role of technology in solving problems … and creating new problems. Like them on Facebook Here and follow them here:

Read more:

Study: WhatsApp Helps Nigerian Women Speak Out

Leave a Reply

Your email address will not be published.