The next wave of machine intelligence

In the first of a new series on the trends shaping the new world of work, our CEO and Founder Charles-Henri Dumon examines the impact of Artificial Intelligence (AI) on the recruitment industry. He sets the scene first, starting with the three waves of AI…

Let me just say first of all that the points I'm making below are based on my own research and understanding of a topic I'm passionate about - they're not meant as gospel!

With AI becoming such a ‘force majeure’ in so many aspects of our lives today, I wanted to talk more about this incredibly important topic, which will have such a big influence on all our lives, jobs and futures.

How Does AI Work? 

For AI to work in practice, training data sets need to be fed into learning algorithms, which are the instructions given by humans to machines to perform tasks. For example, without the input of data scientists or data analysts, computers can’t perform the tasks required of them.

Understanding the Three Waves of Artificial Intelligence 

1. The First Wave of AI: Rules and Patterns 

To help us better understand the nuances, let me talk about the three waves of AI. The first wave doesn’t actually involve any learning, it’s about rules and patterns. Humans establish patterns based on their knowledge and observations, defining the rules to match required outcomes. It would be impossible for humans to define the rules that would enable us to predict the weather, so this would be beyond the scope of the first wave!

Wave 1 is good at perceiving the natural world and at reasoning because you have given AI the rules it needs to follow. It is weak when it comes to learning.

2. The Second Wave of AI: Pattern Recognition 

The second wave is about machines identifying patterns themselves from training data – rather than relying on humans to spoon feed them. As long as you provide enough historic data, the machine can figure the rest out. But it can only learn for a specific goal as each algorithm is only usable for that specific outcome. 

Keeping with the same weather example, if a second wave AI programme was supplied with data for the last 40 years for 500 cities, it would learn the trends and be able to predict the future forecasts based on that historic information. The second wave is where we’re at now.

Wave 2 is good at voice recognition and recognises photos up to a point. It needs a lot of data, therefore requires a high degree of human intervention.

3. The Third Wave of AI: Contextual Adaptation 

Now we move on to the third wave. This deals with the ability to interpret and cross leverage algorithms. Wave three requires far fewer data samples for training, and it can learn and function with minimal supervision. We’re getting much close to a human’s level of intelligence.

Continuing with my weather example, a third AI system could have 10 models each predicting a single aspect of the weather and interacting together to come up with an overall forecast.

Wave 3 is about contextual adaptation and explaining decisions.

Comparing Human and Machine Intelligence 

Daniel Kahneman in his brilliant book “Thinking, Fast and Slow” describes two systems of human intelligence thinking. System 1 is rational, operating within bounded rationality. You look at data and apply your decision based on your experiences and knowledge of the world to make sense of the theory. This is the equivalent of the second wave of AI.

System 2 relies on intuition, which we rely on for making big decisions in our lives. We all have stored knowledge in our subconscious which sees and makes connections, drawing conclusions we can’t necessarily explain. This corresponds to the third wave of AI, which combines the attributes of both systems.

For all the advances that we’re making and from examining the three waves, it’s clear that you can train computers to analyse huge volumes of data by setting out pre-defined rules. And as we know, they’re far quicker and more efficient than humans in doing this!

What we haven’t reached is the third wave. While machines are getting closer to human intelligence, they don’t have the common sense that we have to make decisions the way we make them. Now I for one find that reassuring – you could say that machines don’t have a mind of their own, at least not yet.

For my next instalment, I’ll be looking at the impact of AI on tech and the new world of work with a third and concluding post focusing specifically on AI and recruitment.