Martel attended the European Big Data Value Forum participating in the AI and Big Data public discourse with our unique perspective on AI research and innovation in Europe and beyond. Our take on it? Integrate data-centric machine learning within symbolic AI agent/robotic architectures. But here’s how we came to that conclusion…

Sherlock Holmes’ intelligence is attributed to an uncanny ability to solve complex mysteries starting with very little evidence, and journeying to victory through precise observation and logic analysis. Artificial Intelligence, however, shows us a kind of ‘reverse-Sherlock’, gathering a lot of apparently meaningless data and, without much discernible logic, reaching the correct conclusion and even predicting the future. How is that?

This connection between Artificial Intelligence and Big Data, was the central point of this year’s European Big Data Value Forum. The three-day event, which took place 14-16 October in Helsinki, was opened by Jari Partanen, State Secretary of Finland, who outlined the role of Finland in setting a goal for the EU to become the leading sustainable, socially inclusive and carbon-neutral digital economy, as well as the ongoing national Artificial intelligence programme. On the European Commission’s side, Roberto Viola, Director General of DG Connect, gave the first keynote speech and reinforced the European AI vision by highlighting successful AI applications like autonomous vehicles and further describing data-driven society as going hand in hand with sustainability and environmental responsibility. This idea fuels the vision of Europe’s future: through data, European society becomes better.

Director General Viola also stressed the fundamental role of data for AI, but with the significant addition of high-performance computing (HPC): data enables AI, but both need HPC infrastructure. That is the reason why, in the next political cycle, the European Commission planned and funded them all within a combined vision. The EU is also taking unprecedented steps in addressing and regulating AI, for a human-centric and trustworthy AI implementation that coherently expresses core European values. Part of this effort recognizes that data is paramount, but it’s going different ways according to its source, content, and users; there’s Open Data, B2G (business to government), B2B, and personal data. These differences notwithstanding, it stands without a doubt that data is part of the future of Europe.

The BDVA is an industry-driven association; as such, the institutional and policy-making perspective found an appropriate counterpoint in a series of keynotes from prominent business personalities. A clear recognition of the pervasive opportunity and disruptive potential of AI in most industries invites efforts towards providing business leaders, if not all workers in an organization, with proper training and knowledge acquisition programs about Machine Learning (ML) and AI at large. Training, education, and organizational change management lie at the centre of the challenge Europe must face in order to effectively compete in AI-intensive industries, above and beyond technical and research development, while still preserving its core values of ethics, human-centricity, personal and citizen rights.

The AI factor for European global competitiveness,

While there is still a clear gap with US/China for AI-powered consumer applications, Europe is very well positioned in several industries (manufacturing, logistics, and others) where AI and robotics were adopted more than a decade ago. Also, the amount of data available in most industrial settings, despite its enormous increase in the last years, remains orders of magnitude lower than what can be collected in Web-scale online consumer scenarios.

To really reap AI benefits, industries must combine data-efficient AI expertise with deep and insightful understanding of both business domain and end user. Success stories in manufacturing are never only about AI, nor about AI and robotics per se, but rather about combining AI and robotic automation with additional technological advances within a clear mastery of the business context at hand.

When we add European values and regulations to the mix, such as the AI HLEG Ethics Guidelines, data protection laws and directives, and the need for explainable AI decision-making in some industries, it is not surprising that naïve trust in reverse-Sherlock approaches (successful as they are), is still not gaining too much traction in an European setting.

In industrial and enterprise business scenarios, combining data-driven ML with more symbolic or semantics-based approaches such as knowledge graphs, reasoning and planning aims at killing two birds with one stone, achieving better data efficiency (i.e., less training data is needed) while complying with European ethical and normative rules (e.g., machine learning on a knowledge graph can produce explainable automated advice that still keeps humans in the loop).

The entire #EBDVF19 Forum, while acknowledging the challenges ahead, communicated a great sense of energy and a shared European view of the situation and agenda, bringing together EU institutions, industries, and academia. With such a bubbly and exciting present, it is tempting to resist looking too much farther in the future. Still, the event included a session on ‘Emerging Technologies – Game-changers of Future’. That was the right place to see what could possibly come next in the area of AI, Big Data management, and HPC: from Quantum Computing to various approaches to make reverse-Sherlock work for the real Sherlock Holmes, that is, using data to learn a model that can then be used for reasoning, optimization, and planning in the classical symbolic AI style.

Innovation, in AI and Big Data (as in many other areas of ICT and beyond), blossoms from a complex interplay of different research, business, and socio-political contexts. Here at Martel, as always, we follow, we report, we collaborate, and we make it happen!