How artificial intelligence will change the financial markets
No panic message, no jubilation, but an exciting document for the fintech scene. On 41 pages, the Financial Stability Board (FSB) outlined how machine learning and artificial intelligence will change the financial industry. What makes the report of this international association to monitor the financial industry so special? On the one hand, it shows a realistic view of the effects of machine learning on the industry, far from fear and euphoria. On the other hand, one can read between the lines how hard it is for state regulators to assess and control the new products.
Machine Learning will be the big simplifier for the financial industries, according to the core of the FSB report. That is, because collecting and processing data has become cheaper over the past few years, new players will come into the finance industry and old companies will work more efficiently. That, in and of itself, could first of all have positive consequences.
Not only companies can benefit
Supervisors and companies, for example, thanks to artificial intelligence, can detect scams faster in the future. Long-established banks can better calculate the risk of credit losses. And new companies will use alternative machine learning models to stimulate competition in currency trading or on the credit market. Yes, in fact, consumers could benefit from the new AI boom. Initial companies are trying to better estimate small customers as debtors through a mix of machine learning approaches. So it should be possible in cases where there was too little data for a good analysis, by adding other information, for example from social networks, to assess the collateral. But can the regulators in this beautiful, new world still look after the right? Well, that's where the problem starts.
First, there is the question of how to evaluate the data that feeds the algorithms. How well can firms regulate the valuation of a client's creditworthiness quite differently than with traditional financial data? Because the new openness for small customers with little financial history can mean that massive interference with privacy. Can a company read information from a customer's Facebook profile to rate it? And if so, how much?
And even if at the end of the algorithm "only" uses the place of residence of the debtor to assess their creditworthiness, that may mean that Turkish founders from Berlin-Wedding soon look into the tube - even if they in a normal procedure good chances thanks to a solid concept. At the same time, using new data can create unexpected dependencies.
Artificial intelligence: too little knowledge available
In an ongoing study mentioned by the FSB, the Banco d'Italia, for example, attempts to estimate market sentiment among consumers by analyzing tweets. But what happens when someone tries to manipulate the social network? In fact, in such a scenario, the interest rates could imperceptibly be reduced to one's own advantage, if you trick the algorithms accordingly with bots.
The greatest challenge for regulators, however, are the algorithms themselves. Many models such as neural networks or support vector machines are so complex that it is difficult, sometimes impossible, to see how they come to their conclusions, writes the FSB. These are real "black boxes" that would be difficult to assess and control.
And last but not least, many industry representatives told FSB that it was difficult for government agencies and companies to create audits, "because it requires enough knowledge to understand and monitor AI and machine learning models." Knowledge that - it sounds like that - is apparently still rare.
This is also shown by a response from the Federal Financial Supervisory Authority (Bafin) to a request from t3n.de. Since summer of this year one has a presentation for "financial technological innovations", which is also responsible for machine learning fintech. However, at the moment one is mainly dealing with cryptocurrencies. Only gradually did you want to learn more about machine learning. The background of the members of the unit: mathematicians, economists, lawyers. Programmers and IT experts, however: no indication.
Post a Comment