Will machines learn to act in new situations?

In the foreseeable future, machines are unlikely to be able to effectively operate in the face of such global and unexpected shocks as a pandemic. But with relatively small precedents, they are already learning to cope. Monobank has a whole team that develops AI models for Global Markets. Now she is working on learning the system: artificially creates a large number of precedents for the machine that is out of the general outline. For this, data is synthesized, and various exceptions are generated. By seeing a large number of precedents and adapting to them, the machine will be able to correctly respond to them in the future. Head to bitcoin360ai to check out what they are thinking about it.

How financial market participants can use technology

AI technologies help both individuals and corporate clients trade on the market. But they have different needs and different requests for algorithms.

The development of a competent strategy, which will be based on an understanding of market trends and patterns, is becoming one of the key trading tools in the modern world. In these strategies, the private trader can determine the optimal entry point with minimal risk and the exit point with fixed income. AI stands guard over emotions, formalizing a balanced trading idea for individuals. Find out what to do in such situations by following our link.

Corporate clients, unlike individuals, have more resources to carry out transactions in the financial markets, while they have a request for a larger number and higher speed of these transactions.

However, experts clarify that for many companies such transactions are non-core activities associated with high costs for IT infrastructure, as well as an unjustified risk. The main task of a representative of the corporate sector is to manage the company’s liquidity and currency risk. Sberbank can offer here a specialized solution Trading as a Service2 (TaaS) when the bank provides clients with its infrastructure, knowledge, and trading experience. At the same time, TaaS helps not only directly with managing a currency position, but also with post-trade processes, trade analytics, and data analysis.

Limitations: risks of failure and helplessness in new situations

In addition to the advantages, any technology has its own characteristics, which are important to take into account. History knows many cases when machines made mistakes or turned out to be useless. In 2012, the life of a large American hedge fund, Knight Capital, was cut short literally in 40 minutes due to a technical error: new and old versions of the code were launched simultaneously, the trading algorithm “went crazy” and began to perform unprofitable operations.

Knight Capital lost almost $500 million. In this case, the algorithm failed. It is possible that if algorithms were used in conjunction with AI in this situation, this error would not have reached such proportions: the models would have been able to determine anomalous behavior earlier, and fund managers would have stopped operations.

Actually, it is precisely to prevent such situations in the financial markets that artificial intelligence began to be used. The technology is trained on precedents, and historical models, as a result, the number of errors and failures are reduced.

The peculiarity of artificial intelligence is that the technology is not able to navigate in new non-standard situations. If an abnormal situation occurs in the market, the model is unlikely to suggest the best way out. The pandemic is a prime example of this. The OECD cites that, according to a survey by the Bank of England, during this period, about 35% of banks experienced negative consequences from the functioning of the AI ​​model based on the machine learning method. This is primarily due to the fact that the pandemic has caused a change in many macroeconomic indicators that have become the parameters that are involved in the development of models.

Given these features of AI, many financial institutions do not give it complete freedom of action. For example, in Sberbank, AI does not allow direct control of trading robots. It acts rather as a “smart” assistant and gives the trader recommendations on setting up the executor algorithm. The final decision is always made by the individual. “In this, algorithmic trading is very similar to autopilot. When the flight proceeds normally, the autopilot does an excellent job of controlling. But if the plane flies into the turbulence zone, then the pilots take control of the flight into their own hands,” says an expert.