Amazon's AI Accurately Predicts Sales Of New Products For A Year Ahead
Charles Elkan, head of Machine Learning at Goldman Sachs, announced the breakthrough achievements of Amazon's AI engineers at the annual VentureBeat Transform 2020 conference. According to Elkan, modern machine learning algorithms used by Amazon Web Services are already able to make an accurate forecast for 52 weeks ahead for fundamentally new exchange products.
Unlike analysts, who traditionally rely on historical data, AI "looks" at the current market and consumer demand, processing huge amounts of data. Amazon's neural networks use natural language processing systems to search for similar products, and then study sales trends to determine the potential success of a new product, Venturebeat reports.
When asked whether algorithms can predict stock prices better than traders and analysts, Elkan said: "Sometimes Yes."
According to experts at Goldman Sachs, the key problem of modern AI is bias and bias. Elkan notes that algorithms are not aware of what tasks they perform, so developers must recognize existing biases and eliminate them themselves.
Interestingly, Elkan spoke about the previous version of the AWS stock results forecasting tools, which was updated at the end of last year. Amazon then promised that the new SOFTWARE would provide an increase in the accuracy of forecasts by another 50%, compared to existing systems.
Artificial intelligence is not yet ready to completely replace analysts in the market of exchange-traded investments, but it has already become a good help in certain segments of exchange trading. In 2018, global banks earned more than $ 41 billion thanks to algorithms, and by 2030 this amount will grow to $300 billion, according to IHS Market forecasts. At the same time, millions of Bank employees will fall under the cuts and will be completely displaced by neural networks.