- 62% of companies should use AI in 2018
- AI reshapes almost all banking activities
- Thanks to AI, Nalo automates some tasks with low added value
Genesis of AI
Alan Turing was the first to document the subject in the early 1950s with the publication of an article, Computing Machinery Intelligence, in which he opened the following debate: will machines ever be able to think? Since the publication of this article, interest in artificial intelligence has tested two hypercycles, both of which have resulted in phases of disillusionment. The first phase appeared shortly after the creation of the MIT AI Lab by John McCarthy and Marvin Minsky in the early 1960s, while the second began in the 1990s with Gerry Tesauro demonstrating that reinforcement learning can create game programs that can challenge humans. In each case, failures were caused, in part, by the very high costs associated with building and maintaining these systems and the lack of a rapid return on investment.
Artificial Intelligence: definition and functioning
Artificial intelligence is a scientific discipline related to the processing of knowledge and reasoning, the purpose of which is to enable a machine to perform functions normally associated with human intelligence: comprehension, reasoning, dialogue, adaptation, learning, etc. . Machine learning and deep learning are sub-domains of AI. They refer to techniques for forming algorithms on sets of data, to improve outcomes through experience and learning over time.
So why this time is it different?
In addition to the high costs, the hardware could not keep up with the needs of artificial intelligence, especially to address two key needs: speech recognition and image recognition, which are very demanding in computing power. During the 90s and 2000s the IA was put aside for the benefit of projects focused on raw power and finite element calculations. Today we are at a point of inflection. We have the convergence of algorithms, the influx of data, as well as the computing power. The latter will continue to increase, the algorithms to improve, and the data to grow. The computing power allowed to implement many statistical methods previously inaccessible.
Data, the raw material of these algorithms, are widely accessible thanks to the widespread use of information systems, the internet, mobiles, and other connected objects.
- The development of the Internet has created new needs such as search engines and allowed the implementation of massively distributed architectures.
- The growing needs in robotics, in the space conquest, in the autonomous car, computer security.
- The many commercial applications of the AI crossing machine learning, connected objects or big data.
Little by little, the symbolic steps have been taken. The victory of Deep Blue against Kasparov in 1997 or that of DeepMind, the AI of Google, in the game of Go against Lee Sedol, then world champion. In France, 68 R & D laboratories work on issues related to AI. But the technology has left the research laboratories as well as the universities to settle permanently in the companies. A survey by Narrative Science found that last year 38% of companies used AI and this figure is expected to increase to 62% by 2018.
The AI market will soon reach the momentum needed for widespread adoption primarily through deeper expertise, a broader set of applications, lower IT costs, and support from technology giants like Google or IBM in the US and the emergence of Tencent and Alibaba in Asia.
Artificial intelligence at the service of the financial sector
In the banking sector, where data analysis is an essential issue, artificial intelligence is reshaping almost all activities: trading, portfolio management, risk management, credit, or even customer relations and marketing. . One of the main factors in the evolution of finance is the increasing digitization that is paving the way for AI. As a result, companies in the financial sector have begun to anticipate these transformations.
For example, the US bank JP Morgan has developed a new program called ” Contract Intelligence ” which is able to interpret, through artificial intelligence, commercial loan agreements that required, before the creation of this software, 360,000 working hours a year for lawyers and credit specialists. Another phenomenon, more and more banks invest part of their customer service to these virtual assistants: the “chatbots”. They use AI to mimic human language and respond in minutes to questions from customers who want to know how much it will cost to withdraw cash abroad or if a deposit has been debited to their account. These intelligent systems learn from their mistakes, and represent a cost-free workforce available 24 hours a day, 7 days a week to improve customer satisfaction.
However, when the data is that of the customers it raises questions related to the transparency or the respect of the private life. To meet the growing challenges of securing financial data and the lack of trust of users in the financial sector, FinTech have real assets. The notion of transparency is paramount and the human must keep control of these systems, it must be ensured that he has the skills to work with these new technologies. It seems important to stay in the service of the human and not the other way around. Anticipating this transformation, regulation can be a catalyst for innovation, and the best way to help regulators promote the inclusion of AI is to start developing a set of principles.
In addition to cost optimization for businesses, the automation of low value-added tasks in financial services deeply reshapes the customer experience. FinTech has taken a step ahead by placing innovation at the heart of the customer’s day-to-day operations. Being able to open an account in a few seconds, sign electronically or have a totally personalized advice, all possibilities that maximize their chances of attracting a clientele sensitive to innovation.
Nalo and artificial intelligence
Nalo wants to apply investment best practices by making them accessible to as many people as possible thanks to new technologies. We incorporate all our know-how and knowledge in our algorithms, which have much more computing power and instant that a human manager is unable to implement. To manage the portfolios of our clients, learning algorithms and time series analysis methods are relevant. This is true at a very high frequency, which we do not do, but also on long-term strategies as we are trying to do.
In addition, AI allows us to automate low value-added tasks, thereby reducing operational costs, which translates into lower costs for our customers.
The integration of algorithms in our investment method also allows us to offer equal quality advice for all our clients. Regardless of the amount invested, consulting is equivalent, and human and operational errors are eliminated. Unlike a traditional manager who, by nature, has a hard time getting rid of his own risk aversion, an algorithm has no cognitive biases. As Daniel Kahneman (Nobel Prize in Economics in 2002) shows, the weakness of a manager comes from his emotional bias, and his counterproductive reactions to short-term market movements. Kahneman explains that a human is not capable of developing true expertise in a stochastic environment and that algorithmic rules are more efficient.
Learn more about D. Kahneman: these cognitive biases that deceive us
Paradoxically, the multiplied capacity of analysis of the AI allows us to better know the customer, and thus to offer him a more personalized service. Our robo-advisor provides a tailored response to our clients’ investment objectives. Unlike traditional financial services, which will classify clients on a standardized risk scale (generally from 1 to 10), Nalo creates a unique and evolving allocation according to their heritage situation, the nature of their projects, their horizons. investment or risk appetite.
However, we are aware that the strength of algorithms arises from the know-how of those who make them. That’s why our team consists of wealth management advisors, computer engineers and financial engineers. Thus the financial adviser or the manager does not disappear from the value chain, he becomes the architect. Any intellectual work can not be replaced by an artificial intelligence, this is the case of complex heritage issues (optimization of IR, ISF reduction, transmission of assets) that deserve to be analyzed on a case by case basis.