8 lines
2.1 KiB
Markdown
8 lines
2.1 KiB
Markdown
AI stock price prediction
|
|
|
|
Team 6 did stock market prediction using AI which is unique in the sense that predicting stock market prices has been happening since long before computers. I like to think of the stock market as a decentralized federated learning process that does a really good job of predicting the price of a stock (most of the time). Therefore to beat the large federated process you have to have special information (usually illegal to use) or use the available information better. As one might expect AI systems can be better at predicting stock prices due to their ability to remove emotion and add precise mathematical evaluation. With deep learning based methods having among the best performance across all the papers surveyed by the team. The most interesting avenue that I see is that of sentiment analysis, mainly due to how it adds an order of magnitude more external information to the prediction process. The ethical considerations of stock market predictions are interesting. Particularly from the regulatory stand point as their are various rules in place that are supposed to level the playing field and make it a fair game. Yet large player have an inherent advantage simply due to the scale of resources as well as the actual value they can move around. I see these AI systems simply increasing the power of larger players to outperform retail investors. In the profit maximizing world of stock markets and investing I am unsure that an ethical approach can be found (bar scams and fraud) as it is inherently about beating the other players using whatever means (beside some ground rules).
|
|
|
|
Question and clarifications that I have are:
|
|
- With the ChatGPT sentiment analysis how did this work. Was it along the lines of "Here is a news piece please return a number from -1 to 1". Was there investigation into the accuracy of the sentiment scores given?
|
|
- Will the EU AI Act effect companies using these models to determine customers investments. I.e force companies with the jurisdiction to have "investment explanations" ready to go.
|
|
- Was the ChatGPT sentiment analysis deployed and profited from prior to publication? |