We help customers solve difficult problems in business and industry using artificial intelligence and machine learning techniques, says John Abela.

When did you found Ascent and with what aims?

Joseph Sultana, now Ascent’s CEO, and myself founded Ascent in early 2003. We had discussed the setting up of a software house for many years and finally took the plunge in April 2003. The aim was to set up a software house that would focus on processes and quality.

What elements have contributed to Ascent’s growth?

We believe that the main contributors to Ascent’s success and growth were the emphasis on quality, the adherence to the internally developed processes, and the policy of walking away from projects that are not done right. For instance, we always refused to bid for fixed price software projects unless there is clear, complete, and unambiguous functional specification.

What are the main areas you operate in?

We offer comprehensive software development services mainly to international clients. Ascent is a Microsoft Gold Certified Partner with extensive in-house expertise in Microsoft technologies including SharePoint, CRM, and .Net. We are also involved in postal automation, process optimisation, fintech, digital forensics, logistics, healthcare and, through our special skills team, we focus on the application of AI and machine learning to problems in business and industry.

How much of your efforts are directed to the local and to the international market?

Approximately 90 per cent of our effort is directed to the international markets. These markets include the UK (including the Channel Islands), mainland Europe, North America, the Caribbean, and the Middle East.

What are you doing in the area of artificial intelligence?

We are currently using AI to drive automated trading strategies for the futures markets on the Chicago Mercantile Exchange (Arkenstone), detect suspicious behaviour in e-mails (e-mail forensics), optimise the scheduling of production lines, enhance and de-blur images (car number place recognition), and perform patent prior art. 

Emphasis on quality, the adherence to the internally developed processes, and the policy of walking away from projects that are not done right

In the last two or three years, however, the demand for AI solutions has grown enormously. AI has become truly disruptive. I attribute this surge in interest in AI to the progress made in neural networks – in particular deep learning, which has changed the playing field. With deep learning it is possible to solve, or approximate, difficult problems that were previously beyond reach.

We are currently using AI to drive automated trading strategies for the futures markets, detect suspicious behaviour in e-mails (e-mail forensics), optimise the scheduling of production lines, enhance and de-blur images (car number place recognition), and perform patent prior art.

The media usually portrays AI as taking the form of robots. But AI has much wider applications – can you showcase some of your projects in this area?

The media, and to a certain extent Hollywood, do not help. AI is, in my view, sometimes over-hyped and often misunderstood. AI is not magic. AI algorithms are, after all, just a different type of algorithm.

John Abela

What distinguishes AI algorithms from conventional algorithms is that they, very often, incorporate a learning component. This means that an AI algorithm can learn to perform a task, even a complex one, by learning from training data. The programmer, therefore, does not need to write explicit programming instructions in order to solve a given problem. The AI ‘learns’ how to solve the problem from training examples.

This is a very powerful concept. In the past we have used AI to optimise the scheduling of cabin crew for a major airline, used fuzzy control to monitor the electricity grid to detect anomalies, used a neural network to reduce the energy consumption of large buildings, and used genetic algorithms to optimize the scheduling of manufacturing production lines.

From narrow AI to general AI, how is the journey progressing?

Narrow AI has a specific objective – to develop AI algorithms for a specific task or problem. Narrow AI has been very successfully applied to playing games (chess and Go), automated vehicles, weather forecasting, machine vision and more. It is ‘narrow’ because these algorithms are specific to a one task.

The main objective of AI, however, has always been to build machines that think and reason as we do. It is safe to say General AI is making some progress albeit not as fast as some people expected. In my view this is not necessarily a bad thing. General AI is very exciting but can also be a little frightening. The robots created by American company Boston Dynamics give an indication of what is possible. The warnings about general AI from Stephen Hawking and Elon Musk should perhaps be heeded. Hawking went so far as to warn that, in the wrong hands, general AI could mean the end of humanity.

What are Ascent’s medium-term plans?

Ascent Software currently employs 85 people and we are currently in the middle of a growth spurt. We are recruiting aggressively and will shortly open a new extension to our premises in Luqa. We have a lot of interest for our AI, data science, cloud, and blockchain services and we expect these sectors to continue growing. In the medium term we will also be expanding our offerings in the fintech and digital forensics space.

In IT things can happen very quickly. Technologies evolve and new ones emerge. It is critical to be able to adapt quickly to market conditions and customer demands. Such agility has been instrumental in our success and growth. We constantly keep an eye open for new technologies and paradigm shifts. Deep learning and blockchain are but two examples.

John Abela is a founding partner of Ascent Software and is a director of the company as well as its CTO and COO. Dr Abela holds a Ph.D. in Machine Learning from the University of New Brunswick, Canada and is also a senior lecturer and researcher in computing at the University of Malta. He specialises in machine learning, AI, and optimisation and has worked on the application of machine learning to astrophysics, bioinformatics, image processing, industry and business.