How mathematics will drive the third wave of AI development

From driving your car to using your smartphone to even making a cup of coffee on a coffee maker with a digital timer or logging on to your bank account online, you encounter algorithms in myriad different places today.

Given their ubiquitous nature, algorithms have gripped the imagination of industry, academia, research and development – and almost any sector you can think of. Perhaps the technology most likely to benefit from cutting-edge algorithms – and which in turn, has a ripple-effect on most other sectors today – is artificial intelligence. A key component in the development of strong algorithms is mathematics – the bedrock if you will, to Shor’s algorithm or Schrödinger’s equation. Boolean algebra laid down the foundation of today’s ‘Information Age’. Likewise, Ada Lovelace’s algorithm created way back in 1842, has been widely considered the first computer programme.

As you can imagine, even the smallest AI (artificial intelligence) systems require basic instructions to function. Simply defined, algorithms are step-by-step instructions that help a computer complete a calculation. They are like instruction manuals that allow the machine to know exactly what to do and when to do it. In basic machine learning, therefore, algorithms are the first structural step to building the AI. And practice makes perfect – so through continuously interacting with AI, we can hope to enhance its efficiency.

AI as a technology, if properly leveraged, has endless potential. As highlighted earlier, AI is helping to enhance the healthcare, education, communication, energy, and public safety sectors, and even address critical challenges like global warming.

Today, unfortunately, we are the rudimentary stage of AI and primary  working with simple algorithms such as:

Classification Algorithms
The type of algorithm used to classify a set of data in a specific way.

Regression Algorithms
The type of algorithm used to predict future outcomes based on a set of input data. The computer programs modern meteorologists use to forecast the weather is a good example.

Clustering algorithms
The type of algorithms that take an entire data set and find similarities or differences between specific points. Identifying fraudulent transactions within an accounting document is an instance of such AI at work.

However, to enter the third wave of AI development  – where the intelligence excels at perceiving, learning, and reasoning, and is able to generalise, the need of the hour is revolutionary breakthroughs that will – because of the very nature of AI – originate from the field of mathematics.

We need new mathematics-based tools that can capture the context and structure of the underlying data and are not focused on correlation alone. These tools are primarily shaped on topological and geometrical structures. Today’s AI specialists must focus on crafting efficient algorithms based on these structures that can enable the community to achieve transformative outcomes.

All of this will require new thinking and new ways of learning that compel those working with the technology to go beyond their comfort zone.  As a crucial step forward, today’s digital first societies need to significantly boost the training of future mathematicians and number crunchers. Although AI is attracting more mathematicians than ever, sophisticated mathematical tools that mimic how the brain works and enable us to better structure the logic capability inbuilt in future AI are crucial.

We must accelerate the development of AI technologies and create powerful synergies to push the frontiers of knowledge for the benefit of science, society and global economies.

By Professor Mérouane Debbah is chief researcher, AI and Digital Science, Technology Innovation Institute