Francisco Webber is the CEO and co-founder at Cortical.io, the natural language processing (NLP) startup. The Cortical.io is a computer geek of the old school. He tells Verdict in this latest instalment in our series of our CEO Chat Q&As how he built his first computer with his own hands.

His excitement about technology would push him through his studies at the Medical University of Vienna and through a string of CTO roles. Eventually, though, he grew frustrated with the limitations of the solutions he saw to handle a massive wealth of documents.

In 2011, Webber and his co-founder Daniel Schreiber decided to do something about it and founded Cortical.io. Webber today serves as the CEO of Cortical.io and Schreiber serves as the venture’s CFO. Together, they are developing a new NLP solution to help businesses make sense of oceans of documents.

NLP is essentially what it says on the ti. It refers to the application of computational techniques to the analysis and synthesis of natural language and speech. It is one of the bulwarks of modern artificial intelligence (AI) as it enables modern computers to understand and process the natural language in documents.

Of course, AI developers have achieved a reputation of over-promising and under-delivering on what the technology will be able to do do. The market has taken notice. Developers are not set to change their tack.

“The coming years will be less about making bold statements and more about delivering tangible benefits,” researchers said in a recent GlobalData research report. “Practical uses of AI will be front and center, as businesses ensure they get their money’s worth by using AI to address specific use cases.”

As CEO of Cortical.io, Webber is at the centre of this development. In this CEO Chat, Webber dispels some of the common myths around NLP and his frustrations with the how economics can hold innovation back.

Eric Johansson: Tell us a bit about yourself. What did you do before founding Cortical.io?

Francisco Webber: My career path prior to founding Cortical.io was quite unusual, as I come from both a technical and medical background. I am part of the first generation of computer geeks from the early 60s when people still approached computers with a soldering iron – I had to literally build the first computer I worked on.

I’ve always been interested in information technology and was particularly curious about the correlation between natural science and computing. This eventually developed into my specialty, both in my studies and my career to date – from my studies in medicine through to my experience as a software developer and a CTO, to founding my own company.

Where did the idea come from?

My first encounter with search systems dates back to my medical studies at the Vienna General Hospital. I specialised in blood genetics and serology and participated in various research projects that were heavily dependent on medical data processing. My experience there showed me how difficult and time consuming it is to find relevant patient information in the hospital databases – which I have found to be a reoccurring challenge throughout different stages of my career for different types of information and contexts like patent search and e-discovery. We knew the information was there, but the system was not able to retrieve it.

Out of frustration, I started looking for alternative approaches and followed research done in the field of computational neurosciences, which is focused around collecting data and creating computer models based on the electrical patterns and biological functions of the brain. I was particularly interested in Jeff Hawkins’ Hierarchical Temporal Memory (HTM) theory, which replicates the functioning of the human brain in the space of artificial intelligence. I thought that his interpretation of how information is processed by the brain could be applied to process natural language, so I founded Cortical.io in 2011 to test my idea. The results proved better than expected.

What’s the biggest misconception people have about NLP technology?

What people often get wrong when it comes to NLP is assuming that language is a statistical phenomenon. This is a habit that scientists and academics have adopted from research areas where we might know the effects of a certain action or the way to manipulate it, but still haven’t understood the real principal behind it.

In those situations, the research relies on a statistical model, which would be able to identify that there is an issue, for example, but won’t be able to give you a suggestion on what the issue might be and how you’ll be able to solve it if there isn’t enough historical data available to statistically inform that.

The truth is that NLP is not about statistics – it’s about context, and about enabling computers to understand the context of text and spoken words in the same way human beings can.

What one piece of advice would you offer to other CEOs?

I’ve always been a very practical person. Everything I have done in my career has been focused on practicality and my advice reflects that. I think it’s really important that CEOs take time to really understand the majority of the business’ day to day processes and tasks. From product development, to sales, customer relations or marketing, the more you understand what your team is facing, the better for the business and for your leadership skills as well, as you’ll have the knowledge and understanding needed to make the right decisions.

What’s the most surprising thing about your job?

The most surprising thing for me in my early days as a CEO was how long things take. And I now understand why. Unlike the academic world, the business world doesn’t have an end point of satisfaction where you’ve found the answer to your question, or the solution to your problem. When it comes to business, you’re working towards an ever-evolving goal, the success of which depends on a whole package that needs to be constantly updated and adjusted to make a real impact. There is no end point in business and good things do take time.

What is the most important thing when you want to scale a company?

I think it’s crucial that you, as a leader always have plan B and C. It’s about having a few parallel strategies in place that can help you turn potentially negative outcomes into positive ones for your business and your team. In the best-case scenario, those strategies would even complement each other, and in the worst, you’d just need to switch one strategy for another and keep moving.

Some people, especially in the start-up world, may argue that when having a few strategies in place you may lose your focus, but for me being solely focused on achieving one outcome is like standing on one leg – you may easily lose your balance.

What’s your biggest pet peeve?

How hard it is nowadays to not only bring innovation to life but to also sustain it outside of big structures. Nowadays, innovation is mainly financed by businesses, hence it must follow the same patterns of scalability and growth. Investment in new ideas depends on the revenue expectations, not their geniality. I prefer not to think of all the great ideas that have disappeared because there was no immediate business case. As a serial entrepreneur and passionate researcher, who is always trying to stretch the limits of what’s possible, I believe more should be done to foster innovation outside of the business world.

What’s the strangest thing you’ve ever done for fun?

The strangest thing I’ve done for fun is probably my job. Just like many kids, I hated grammar at school and thought I would never need that in my life again, but here we are. If anyone told me back then that I’d be building a business and providing for my family using grammar, I’d probably laugh at the idea.

What’s the most important thing happening in your field at the moment?

At the moment, the biggest thing when it comes to natural language processing and AI in general is efficiency – and the fact that the industry has somehow forgotten about it. There is a big disconnect between what researchers are working and focusing on and what the world needs right now.

The problem is, that similar to the way Bitcoin mining works, AI productivity is directly linked to the amount of energy that is pumped into it. Therefore the industry has reached a point where it depends on an environment and technology that can only be scaled and improved with larger amounts of energy. And that’s not very efficient.

As a business, we’re now actively looking to address this problem by focusing on improving how we’re making NLP more efficient from both a hardware and software perspective. And while we’re at a very early stage of those improvements, we’re already seeing 4,500 times more energy efficient operations as part of the trial phase.