Musings after Amsterdam (and the World Summit AI 2018)

A couple of weeks ago I attended the World Summit AI 2018 in Amsterdam.

Coming from a non-technology background (I originally trained as a medical doctor) and as someone who had since worked as a management consultant and more recently have been working with various startups (including ConscientAI), I was keen to understand more about AI from a business point of view.

WSAI 2018 was an amazing gathering of more than 6,000 attendees from across 160 countries who are enthusiastic about the potential of AI to solve various challenges. Thinking back at the various speakers, workshops and conversations over coffee, I wanted to share my thoughts and learnings.

What are the current trends?

Democratisation of AI was a theme that came up several times and was reflected in the positioning of the offerings from the major technology providers that had a presence at the summit including Google, Microsoft and IBM. The message was clear — don’t spend your time and resources trying to rebuild the engine when robust, scalable and secure engines can be accessed via APIs and other tools. The effort of the majority, especially from a business point of view, should be on using the domain expertise on defining the actual problems and fine tuning the engine to solve those problems. However, I don’t believe this means that expert technology knowledge is not required or that these tools will be suitable in all scenarios, for example when extended / specific functionality or flexibility is required. It is important to keep investing in AI research and to improve the core AI technologies to continue the progress made to-date.

A related point was around the importance of technology interpreters. Many companies are investing heavily setting up innovation teams and individuals that have a dual skill set. First, a good understanding of the industry domain and how to identify and frame business problems. Second, a broad understanding of technology and AI to be able to have a sensible discussion with the AI engineers to ensure the models are solving the real business problem. We at ConscientAI are working on a fashion analytics platform and it reinforced the importance of getting the domain expertise to ensure your solution is relevant and important to the industry.

Another important trend was the emergence of Cloud AI, which is where the heavy lifting and processing required for AI is done centrally allowing a number of smart devices to have much more functionality. This was especially exciting thinking about the many applications in the healthcare space — including devices that help people see better, and artificial limbs that help people lead a more normal life.

How do you get the best value from AI?

The oil that make the AI engine run is data. Similarly to oil, data can be of different grades and this has a huge impact on the outputs and therefore the value that AI adds. We know from the experience on various projects at ConscientAI that this is often an area of challenge, whether it’s having access to an adequate amount of training data to issues around data governance. Securing the data supply chain and thinking about changes to data availability (and how models may need to modified accordingly) will be an increasingly important future strategic consideration on AI initiatives.

Similar to other technology projects, testing and implementation are very important elements (and perhaps overshadowed a bit by the more exciting model building). Due to the nature of AI, especially deep learning based techniques, it is often difficult to document exactly how a model derives the outputs. That doesn’t mean to say controlled testing shouldn’t be done to validate the outputs against expectations within a number of controlled scenarios. This would be especially important and will be enforced legislatively in industries such as financial services and healthcare — indeed, I was speaking with someone who was involved in developing an AI based solution for a hospital and was going through CE accreditation required for medical devices. As with other transformative programmes, implementations should be mindful not to disrupt Business as Usual and incremental implementations of functionality enables fine tuning and risk reduction.

AI will impact the many job roles and the way we work, there is no doubt about this. Having read a lot of scaremongering about the negative impact that AI will have on jobs, it was refreshing to understand it from a more balanced point of view. The overwhelming view was that AI will not replace humans but will supplement their capabilities. It was also discussed how AI could automate the more mundane elements of jobs enabling humans to focus on the more complex and therefore more rewarding elements, leading to increased job satisfaction. Having had many discussions on this in the healthcare space, in the context of AI powered solutions such as Babylon and what place they have in healthcare systems like the NHS in the UK and how acceptable they would be to doctors, I know that it won’t be that straight forward. However, I believe that we need to have an open discussion that is based on the facts — that will be the only way to work through what is undoubtedly an area that is going to stir up many emotions.

What does the future hold?

The skills required in the field of AI are evolving with some convergence of core technical and business analysis skills. For example, Cassie Kozyrkov talked about Googlers being trained in Decision intelligence — a mixture of applied AI, machine learning, statistics and analytics skills for solving business problems.

More thinking and focus on governance for AI — I think there are several elements to this. First, there will be a huge focus on the data that feeds the AI engines and how that AI is used to make decisions that affect our work and our lives (i.e. medical diagnoses, decisions around insurance, etc). Second, there will be checkpoints and assurance frameworks to ensure the validity of the outputs. One of the advantages (or disadvantages when the output is incorrect) is the ease by which AI can be scaled — we need to be sure that it works at a smaller scale before amplifying the issue at a larger scale.

My son started school last year and I sometimes find myself thinking about and trying to predict what I need to do ensure that he is in the best place to get the right education and skills. From my own career experience and what was reinforced at WSAI, it feels like the focus is shifting away from specific (technical) skills and from developing a specific career (planned from step 1 through to 10). It feels like some of the more generalised skills related to critical thinking, lifelong learning / flexibility and curiosity mixed with a healthy dose of creativity (not of the traditional artistic nature necessarily) are going to be key determinants of success.

In summary, WSAI 2018 was a very interesting event that has made me think more about the upcoming trends in AI and what we need to do remain relevant. Lots of conversations to follow up on and the Colombo chapter for City.AI to establish!

Feel free to drop me a line at — always happy to chat.