ficial Intelligence for Beginners //

Everyone is talking artificial Intelligence. It’s easy to understand — there are now free or inexpensive tools that can create AI-generated text and images in a wide range of styles and in a matter of seconds.

It’s certainly exciting.

Stop for a minute and ask yourself some questions.

• What is AI?
• How long has it been around?
• Do I understand the differences between AI and Machine Learning?
• What is deep learning, and what does it mean?

This article is not for you if you answered yes to all of them. Read on if you questioned some of the questions.

The AI revolution begins… now?

Let’s begin by describing some of the background.

AI is a new concept?

No. AI is at least conceptually as old as 1950 (more about that later). In the 1960s, and 1970s, AI became more practical as computers became cheaper and faster.

AI marketing – is it a new concept?

No. AI is not just about creating content. AI has been used to power product and content recommendations for many years. AI has long been used in predictive analytics, which is used to predict behavior based upon large datasets.

Since almost a decade, well-known vendors have incorporated AI into their products. Adobe Sensei and Salesforce Einstein date from 2016. Oracle has been involved with AI for at least the same amount of time and probably even longer; they just didn’t give it a cute title. Pega is another veteran AI user, first using it to predict the next best actions in its Business Process Management offering and then in its CRM platform.

Is generative AI a new concept?

Conversational AI. Conversational AI. AI writing tools. All the current phrases, with overlapping meanings. Generative AI creates text (or images or videos). Conversational AI creates text in response to a human conversation (think AI powered chatbots). AI writing tools are designed to produce customized texts when needed. All these solutions rely on “prompts,” that is, waiting to be asked a specific question or given a task.

All this is new? No. Its wide availability is what’s new. Since years, natural language processing (NLP), and the generation of natural language (NLG), have existed. The first is AI-powered text interpretation; the second, AI-powered text creation. According to my own reporting from 2015, AI-powered NLG created written reports for doctors and industrial operations, and even generated weather forecasts for Met Office, U.K.’s national weather service.

Text out, data in. Not as widespread as ChatGPT.

Video too. AI had been used by 2017 to create personalized video content, but also individualized video. The video is generated so quickly that it looks like the video is streaming. It is not widely available but rather a cost-prohibitive enterprise service.

Dig deeper:


ChatGPT

Guide for Marketers

What AI is: The simple version

Let’s start at the beginning.

Start with algorithms

A set of rules that are followed by computers to solve problems or complete tasks can be described as an algorithm. Does “algorithm’ come from Greek? It’s not. The word comes from a part of the Arabic name of an Arab mathematician who lived in the 9 ^th century. It doesn’t really matter.

It is important to note that AI is not the same thing as using algorithms. Let’s look at a simple algorithm. Say I’m running an online bookshop and I want to provide product recommendations. I can create a hundred algorithms and train my site to follow them. If she searches for Jane Austen also show her Emily. “If he search for WW1 books also show him WWII books.”

My volumes of detective novels will need to be tagged, of course. But so far it’s been easy. These are, on the one hand good rules. They are also not “intelligent”. This is because the rules are set in stone until I change them. The rules won’t change if people who are searching for WW1-related books ignore WW2-related books. They continue to do what they are told.

If I had Amazon’s resources, I would make my rules smart — that is, they could change and improve based on user behavior. If I had Amazon’s share of the market, I would have an abundance of user behaviour that rules could learn from.

AI is when algorithms can learn themselves, with or without human supervision.

But wait. But wait.

AI versus machine-learning

Purists do not consider AI and machine-learning to be the same. The terms have become so interchangeable that it is impossible to go back. When people are talking about pure AI or AI in its original meaning, they use the term “general AI”.

I warned you that we’d go back to the year 1950. Alan Turing, a brilliant computer-scientist, was one of the most influential figures in history. His code-cracking work helped the Allies defeat the Nazis. He was treated horribly by British society because of his (still illegal at the time) homosexuality. This treatment led to an apology from Gordon Brown more than 50 year after his death. You deserved much better.”

What about AI? Turing’s landmark 1950 paper “Computing Machinery and Intelligence” is widely known today as the “Turing test”. It proposes an artificial intelligence (or machine) criterion. We can attribute intelligence to a machine if a human interlocutor cannot distinguish between the responses she receives from a machine versus responses from another person.

Turing’s proposal is certainly not without objections (and Turing’s test isn’t even cleverly designed). This was the beginning of the quest to duplicate — or at the very least, create the equivalent — of human intelligence. IBM Watson is a continuous pursuit of this goal (although there are many less ambitious and profitable uses cases).

Nobody believes that a machine like Amazon’s product recommendation system or a ChatGPT content creation engine are intelligent the same way humans can be. They are incapable of knowing if they are doing the right thing or not. Instead, they rely on predictive statistics and data to make their decisions.

All the AI that is discussed here is actually machine learning. We won’t stop anyone from calling it AI. There are many reasons to believe that “general AI” or human-level AI is not too far away. You can read Erik J. Larson’s ” the myth of artificial Intelligence: Why computers cannot think like we do.”

What is ‘deep Learning’?

You may also hear “deep learning” used in relation to AI. Is it a different concept from machine learning? It is. This is a huge step above machine learning. The AI can now detect patterns, and handle images and video as well as numbers and words. Here’s a short version.

Deep learning relies on a network of artificial neurons, which is a layer of artificial bits of math that are activated when an input occurs, communicate about it and then produce an outcome. The nodes are able to determine the accuracy of the output, as in traditional machine-learning, and adjust accordingly. The neurons are then trained by “back propagation”.

There are also “hidden layers”, which multiply between the input and output layers. Imagine these layers being literally stacked: This is why machine learning of this type is called “deep.”

The stacking of layers in a network is much more effective at recognizing patterns within the input data. Deep learning is useful for pattern recognition because each layer breaks complex patterns down into simpler patterns.

Do AI vendors exist in the martech industry?

What you mean depends on what you are saying.

Vendors using AI

It is estimated that over 11,000 vendors are in the martech industry. Most of them (or at least most can argue that they are using AI) use AI. They don’t use AI just for the sake of it. They use it to accomplish something.

• Create commerce recommendations
• To write email subject lines.
• Recommend next-best actions to sales representatives or marketers.
• Chatbots powered by the cloud.
• Write advertising copy
• Create content for multivariate large-scale testing.

Lists are endless.

My point is that AI can be compared to salt. To make food taste better, salt is added. We all like salt when it is used properly. Who says “I will have salt for dinner” or “I want a snack, I’ll take some salt?”

We add salt to food. We use AI in marketing. Salt and AI can be used together for research, but not much else.

There are many martech vendors who use AI. Are there martech vendors that sell AI as a standalone product?

Vendors selling AI

In the martech industry, there are very few. AI is a term used to describe AI software that engineers have created and can be integrated into other solutions. There are many engineering companies that sell AI software. However, they tend to target IT departments rather than marketing groups, and the AI software is primarily used to support back-office functions rather than sales or marketing.

One or two companies clearly target their products towards marketers. It’s not enough to make a category populous in the marketing technology landscape.

The surface is just scratched

This article will only scratch the surface of a topic that is incredibly complex, with a rich past and an uncertain future. Of course, there are ethical issues to be addressed, including the inevitable instances where machine learning models would be trained using biased data, and the equally inevitable plagiarism of human content by generative AI.

This should be enough for the moment.

The post Artificial intelligence: A Beginner’s Guide appeared initially on MarTech.