Artificial Intelligence is the source of many questions and fears. What is it really?
There are two types of Artificial Intelligence: strong and weak. Strong AI is the one capable of replacing mankind – that is to say – it can do everything, understand everything, and perhaps even have a conscience. This is the AI that forms the foundation of much science fiction literature. It is now being researched, but it is still theoretical and does not yet exist. At BNP Paribas, we aren’t interested in such research. The AI we are interested in is that used everywhere today, across all industries: “weak” AI. It is capable of accelerating and automating time-consuming and repetitive processes.
How does this "weak" AI work in practice?
It is based on two techniques. First, supervised learning, which involves training a machine to perform a specific task by showing through examples and feedback how a human would do it. The second method is unsupervised learning, which means that the machine is able to learn by itself from data provided.
To illustrate these two approaches, one can observe the evolution of the famous AlphaGo software, developed by Google to take on champions of the highly complex game of Go. Until 2017, AlphaGo was trained using data from games already played, i.e. supervised learning. Then an unsupervised version of the software was unveiled, AlphaGo Zero. Knowing nothing more than the rules of the game, this new version played millions of games against itself, before winning every one of its official games.
Photo : Hugues Even
What are the benefits of AI?
Currently, AI allows for automation of repetitive and basic tasks that do not require human intervention. Specifically, the idea is to allow employees to refocus their time on what they can do best. At BNP Paribas, for example, AI frees up time for our operators in the back, middle or front office so that they can be fully available for their exchanges with the client, performing complex analyses and making decisions. This pushes us to ask: “How can AI improve on what humans do best?" or "What organizational changes are needed to accommodate this new form of human-AI collaboration?"
What exactly is BNP Paribas' strategy for AI?
We base ourselves on the principle that AI allows us to respond to the small problems of everyday life. No attempts are being made to develop projects on a colossal scale – projects of this type never end. Instead, we focus on automating daily tasks that meet the pragmatic needs of our clients.
Our AI is taught to master basic functions, such as text, image, and speech recognition. Our solutions are developed as web platforms and API, which can then be used by all the teams in the Group. We have a very agile way of doing things leading us to review our traditional IT infrastructure. We are moving from a monolithic, closed IT, in which we sought to do everything, to a more modular, open architecture, making micro-services available in the form of applications.
Our AI is taught to master basic functions, such as text, image, and speech recognition.
We have a very agile way of doing things leading us to review our traditional IT infrastructure.
What concrete applications within the bank come to mind?
We are especially proud of our intelligent translation engine. As a global bank, we regularly need to translate confidential documents, e.g. contracts for our clients or emails. However, the content is often too sensitive for public tools such as Google Translate. The vocabulary is often also very specific to banking. Our internal translation engine can translate content from up to 15 languages. The tool allows us to preserve one of the primary interests of our customers: protecting their data, while producing translations that meet our expectations.
There are also many other use cases. Some of our teams need to extract tables from annual reports in order to comment on them. We have developed a solution which is able to extract that data and then to produce related comments (through Natural Language Generation or NLG). Detailed and substantive analysis is then provided by our experts.
Finally, a third project we are very proud of is our internal search engine, called SEARCH. Large organizations, such as ours, have many internal search engines, but often lack a single search engine, capable of making all knowledge available in one place, like Google. Well, we’ve done that! We created a global index, accessible to all, which is able to manage jurisdictions and accesses. In the long run, it will become a major lever for the transformation of the bank.
“ We created a global index, accessible to all, which is able to manage jurisdictions and accesses. In the long run, it will become a major lever for the transformation of the bank. ”
Chief Data Officer chez Corporate & Institutional Banking (CIB)
What is the impact of AI on the way a company operates and grows?
AI can completely transform the operation and culture of a business. This is why it is important for us to start small and grow step by step. AI has been generally very well received by employees as it is seen as a mean to help save time and do work more effectively. For example, thanks to word of mouth, the translation tool was quickly adopted by thousands of users, without much communication effort.
What are your key next steps?
As far as the banking world is concerned, I think Compliance will offer the greatest opportunities in terms of AI. Compliance covers an immense scope as it touches all the activities of the bank making it a prime opportunity due to the large amount of unstructured data it generates. As regulatory scrutiny continues to increase, it is more and more crucial to look for optimization in Compliance.
The banking industry has become a knowledge industry and its future lies in its ability to manage this knowledge. This means connecting all its data, so that the bank can generate intelligence, designed to better serve clients.
Chief Data Officer – An increasingly important role
Hugues Even is Chief Data Officer of Corporate & Institutional Banking (CIB). He explains his role as follows: "It is to upgrade, protect, value and make data available to our internal or external stakeholders to optimize some of the tasks performed by the bank."
Within CIB Analytics Consulting, the Artificial Intelligence Lab’s purpose is to deal with unstructured data such as text, images or speech. This type of unstructured data represents 80% of the overall volume of data – a huge potential to be tapped into. The Lab consists of data scientists and developers. Their goal is not to conduct research confined to the Lab, but to develop projects that will be rolled out and be of real use to clients.
In parallel to the Lab, Hugues manages the Central Data Office. Within CIB, each business and region has its Chief Data Officer. Hugues' role is to work hand in hand with them to deploy policies, procedures and control plans to ensure compliance with regulations – such as GDPR – and data quality.
Crédit photo : header ©Gorodenkoff