Meet StockGPT, the ChatGPT of Stocks

OpenAI’s popular offering ChatGPT has inspired a host of AI-powered tools such as Character GPT, Claude and Chat BCG. And then there’s one for stocks, called StockGPT. StockGPT is free to use and can be accessed here. 

StockGPT is an AI-powered financial search tool that has been trained on all of Tesla’s quarterly earnings call transcripts dating back to the second quarter of 2011. The prototype was announced on Twitter by Alec Corum, who is part of the team working on developing the tool.

StockGPT can be used to find commentary on specific topics for a certain quarter and can also be used to analyse performance over a certain period. It also gives insights into company struggles and even tells you bad jokes.

Presently, the prototype tool is only designed to answer questions based on the context of the specific dataset (Tesla’s quarterly earnings call transcripts). For any question beyond this context -- the tool responds with “I don’t know”. The platform cites explicitly that the tool is still in an “experimental state", specifying its responses can at times, be inaccurate or completely wrong. The site also provides a list of examples of past mistakes the AI tool produced.

Corum added that they would add additional companies / features over time once they refine the product.

One of the users who tried the prototype asked Corum on Twitter about how expensive it would be to get the transcripts and train the model on the all 500 S&P companies to which Corum said that a rough estimate would be $700- $1000. He also explained that it is easy to get all the transcripts but they are using semantic search and thus have to embed every single transcript that exists.

Corum said that he hopes to support all the companies, provided they can reduce their cost of training and inferencing as it is skyrocketing.