Summary
In the Season 1 Ep.10, Mike Schuster, Managing Director at Two Sigma, discusses the use of artificial intelligence (AI) in finance. At Two Sigma, AI is used to predict future stock prices and to assist in making informed investment decisions. The AI system relies on processing a lot of data, including volume, prices, news, international politics, weather, etc. Two Sigma views their work as similar to that of scientists. The guest speaker learned about neural networks while working on speech recognition with a professor in Japan, and he taught himself about neural networks by reading books and programming with a small group.
Shorter-term trading decisions are easier than longer-term ones due to the amount of data available. The decision between focusing on interval predictions for a week versus a shorter period is based on the amount of data available. Two Sigma is a smaller company than Google, with around 1,500 employees based mostly in New York. The culture at Two Sigma is more traditional and focuses on doing everything right, likely due to the field they work in, where mistakes can be costly. Employees at Two Sigma dress casually and work from normal hours.
The guest discusses using news articles to predict stock market movements. They mention that getting a sentiment out of a news article is a possible way to predict a stock market movement. They acknowledge that sentiment analysis can be noisy due to the many factors that influence stock prices and company evaluations, such as mergers and acquisitions. Decision-making in finance based on historical data is challenging because the data are finite, and there is a lot of data that may not be reliable, making it difficult to select the most pertinent information.
Two Sigma is a finance company that deals with big systems and a lot of money. AI can also be used in finance to determine outliers and improve transparency of the system. Reinforcement learning can potentially be used to make decisions for trades, but there are complications like risk, regulation, and the need for transparency and explainability. The field of finance is gradually shifting towards using more complex systems and models that can be improved through machine learning.
While accurate predictions about how finance will change due to AI in the next 5-10 years are difficult, more parameters in models will be trained through learning from data systems such as maximum likelihood and reinforcement learning. The gradual shift to using machine learning in finance will help investors make better investment decisions, resulting in improved outcomes. To keep himself motivated and sane when working on difficult projects, Mike suggests not giving up easily and thinking about things that will happen in the future. He also recommends having other things going on in life, doing sports, using scientific common sense, and looking at things from different angles.
