Happy New Augmented Intelligence Year – 2019

While 2019 marks the official launch of our company, Nebuli – in the good old town of London – 2019 also appears to be the year where the artificial intelligence (AI) hype (amongst other hypes – Blockchain anyone?) might start to die away. And die away it must! I say this despite being an entrepreneur who has been involved in the AI, tech and media sectors since 2005. But I’m old enough to remember the .com hype that led to disastrous losses, not just for investors, but also in market confidence.

My team and I at Nebuli are passionate believers that AI can bring significantly more positive outcomes to society than some people believe. So much so that we doubled down on Nebuli’s foundation, pushing for the lesser-known concept of Augmented Intelligence which is far beyond the current model of artificial intelligence. Why do we do it? I promise you, we are not aiming to start yet another tech hype, but rather start working on and debating about the true realities of AI (both good and not so good) that we are seeing today and the reasons why we are witnessing the need for augmented intelligence.

First, though, let’s start with the basics:

A.I. vs A.I.

Errm, which one is which you ask? Well, that’s the point! Mathematically and algorithmically speaking, there is almost no difference between artificial intelligence and augmented intelligence. The key outcome from whichever model or terminology is to leverage technology to augment human intelligence and productivity. Simple, right? Well, it could’ve been, but it is not that simple because the biggest challenge that faces any growing technology is its application. In other words, connecting this technology’s concept with the real world. 

To date, the vast majority of the AI hype (i.e. artificial intelligence) has been focused on automation and robotics. Don’t get me wrong, we think this is a great achievement considering that the concept of AI has been researched and developed since the 1940s. However, our argument is that we need to start looking beyond automation, not to mention the other hype that AI will “replace humans” in almost anything these days. Or, that robots will take over the world. Dudes, stop watching The Terminator movies! This will not happen unless we explicitly allow it and program it to happen.

Let’s actually look at the problems we are seeing with the AI market today and and aim to solve them in order to bring the positive outcomes I mentioned above.

The Hype Problem: Overpromising and Underdelivering

Our aim here is to positively critique the market in order to ensure that we, the AI community, collectively deliver the positive change that we all aim to achieve. However, we believe that 2019 will be the year where AI weaknesses and overhyped promises will be exposed. We have already seen reports emerging highlighting underdelivered promises in major projects. One such report from MIT Technology Review pinpoints this very issue with IBM Watson.

The overuse of the keyword “AI” can be further witnessed in companies applying almost no “real” artificial intelligence, or some minuscule automation within their workforce. According to a recent report from Crunchbase (published on January 11th, 2019), there’s no market consensus over exactly what counts as an AI startup and that startups are raising massive sums of money off the AI buzzword. Furthermore, the author concluded from their data that there are startups who overstate their proximity to AI. 

There has also been a research study reported on Mindzilla a few days earlier revealing “severe limitations” of deep learning machines and that they “have a long way to go” before achieving key promises made by some leading players in the field.

This is, I’m afraid to say, the “.com bubble” all over again, but we hope it will start to die down in 2019. While I believe most of these promises were well-intentioned, the reality is this: we must consider a different approach of applying AI (in whatever form) in real-world applications that are less expensive, more accessible and, above all, more relevant to people’s needs. Let’s tame down the promises and start delivering more!

“Top-to-Bottom” Approach Could be the Key Problem

The most common implementation of AI-based systems tends to be, what we describe as, “top-to-bottom” approach. This predominantly involves an organisation developing an AI system with vast capabilities, with endless libraries and then try and work out different applications within specific markets. In this scenario, the market would be required to develop additional layers and new API libraries to customise their existing systems and make them “AI-ready” in order to take advantage of the above AI offerings. With this, as you may expect, comes major time and financial commitments for customers, not to mention, miscommunication between parties and inadvertent overpromises.

We fundamentally disagree with this model where, essentially, the markets interested in applying AI tools need to significantly change their existing systems to apply somewhat generalist AI applications. As a colleague of mine put it to me simply: “it’s like trying to fit a big square into a tiny circle.”

We at Nebuli believe in the exact opposite model. A Bottom-up approach where we do not expect markets to make major (if any) changes to their existing systems in order to take advantage of Nebuli’s capabilities.

Instead, Nebuli’s structure is the one that adapts to specific market needs, right to the basic details. For example, a legal firm may not need the backing of a huge generalist AI backbone that can be applied in other markets, but rather it would require a smaller and lighter, yet highly specialist AI tool that performs the specific tasks this firm needs. And, without any need for this firm to develop additional layers and/or libraries to connect its in-house software with this AI system. This saves customers thousands of man-hours of redundant development work, and of course several thousand in costs.

This is why we refer to our approach as augmented intelligence, where it is not focused only on having intelligently indexed large datasets from all sorts of markets and creating endless libraries for all sorts of tasks. Instead, building a more specialist approach that is faster, more relevant and more understanding of the needs of the users and being able to adapt accordingly. This is more of a personalised AI that inherently makes it smarter and goes far beyond automation, which indeed also does require a high level of intelligence. But having the capability of understanding relevance and impact requires a much more augmented model of intelligence. Dare I say it, more human.

We believe this is the future of AI in the coming years as markets have started to demand more relevance from AI providers at lower costs, with better interoperability and, above all, offering better data security. Our core vision and current work with Nebuli are to offer all of these capabilities in a single package and with a plug-and-play model. Will this be the revolutionary model for future AI applications? As the entrepreneur and cofounder behind this company, I say “yes” (with a cheeky grin, while sipping my coffee with my pinky out). But my honest and fair answer is “time will tell”. But make no mistake, markets all over the world have become much more aware and educated about AI in terms of its opportunities, capabilities and, above all, its weaknesses. The next 10 years will be an exciting period for augmented intelligence, I can assure you that.