This paper conducts a zero-shot analysis of ChatGPT's capabilities in predicting stock market movements using multimodal data, including tweets and historical stock prices. The study positions ChatGPT as a 'Wall Street Neophyte,' highlighting its limited success in this domain. By evaluating its performance without fine-tuning, the research compares ChatGPT against both state-of-the-art methods and traditional approaches like linear regression, revealing that it underperforms across these benchmarks. The analysis also explores the impact of Chain-of-Thought prompting and the integration of social media sentiment, but finds that these strategies do not substantially enhance prediction accuracy.

The findings underscore significant limitations in ChatGPT's application to financial tasks, particularly in terms of explainability and stability, suggesting that specialized training or fine-tuning may be necessary for effective use in stock market analysis. This work provides critical insights into the current capabilities of large language models in finance and serves as a foundational study for future research aimed at leveraging social media data and advanced prompting techniques to improve financial prediction models.