Written by Fabio Colasanti, Head of Data, Data Language
In my experience at Data Language, an AI product scale up that works across the USA and Europe, there are three key steps that Europe could take to become a global leader in AI:
- Let go of the big bang approach
There is still a tendency globally to approach AI with a waterfall methodology; working up huge specs that everyone is happy with, for programmes with many varied features, and then aiming to launch the finished product only when all those features are ready to go live
There are two aspects of this big bang monolith tendency in particular that are holding Europe back: long procurement cycles and vendor lock-in, which severely limits business agility; and long delivery wait times caused by the complexity of the project and the desire to launch the finished product with all functionality (the big bang moment).
Europe has the chance to get ahead by embracing the lean start-up approach, “launch it quick, launch it early”. Instead of wrapping up multiple process responsibilities into a single, complex, monolith application with too broad a focus, European AI should aim to deliver simple scopes quickly.
- Follow technology patterns
Businesses should treat AI and machine learning as they would any other software engineering project and use contemporary agile engineering patterns. First, automate as much as possible from the outset, including continuous integration and deployment and test-driven development. Second, understand your predictive analytics scaling needs and design with this scale in mind. There’s no need for Europe to reinvent the wheel – instead we should embrace contemporary deployment architectures such as Kubernetes and Containerisation. Third, great data scientists and great software engineers shouldn’t be an ‘Either / or’. To develop market leading AI you need both, so either ensure that your data scientists are themselves excellent software engineers, or pair them with people who are.
- Commit to collaboration
For Europe to become a market leader in AI, more steps need to be taken to promote SMEs and help larger companies move faster. There is already excellent work being done in the UK, for example, Digital Catapult’s machine intelligence garage, which matches corporates with innovative scale-up AI products, and across Europe too, with the Horizon 2020 SME instrument calls, but we need more to truly stand apart in the field. Only by connecting the innovation and agility of smaller companies to the resources of enterprise-scale organisations, will we reach Europe’s full potential. SMEs and start-ups can prototype to their hearts’ content, but without industrial use cases, the challenge will be demonstrating value.
This also applies to ethics.Without solid ethics foundations, AI projects will struggle to deliver predictive analytics that can be trusted, which will, of course, hinder adoption, use cases, and market leadership. European companies developing AI must understand data ethics, and how it affects the entire ecosystem that their predictive analytics solution will be part of. This should encompass everything from bias, and to transparency and explainability.
There are good frameworks coming out, but data scientists and software engineers need more practical guidance; clear steps to take to ensure that their projects are built on ethical foundations. In order to achieve this, cross-industry and governmental collaboration is required, and codeable ethics frameworks need to emerge.