Will Affective Computing Emerge From Foundation Models and General Artificial Intelligence? A First Evaluation of ChatGPT

ChatGPT has shown the potential of emerging general artificial intelligence capabilities, as it has demonstrated competent performance across many natural language processing tasks. In this work, we evaluate the capabilities of ChatGPT to perform text classification on three affective computing problems, namely, big-five personality prediction, sentiment analysis, and suicide tendency detection. We utilize three baselines, a robust language model (RoBERTa-base), a legacy word model with pretrained embeddings (Word2Vec), and a simple bag-of-words (BoW) baseline. Results show that the RoBERTa model trained for a specific downstream task generally has a superior performance. On the other hand, ChatGPT provides decent results and is relatively comparable to the Word2Vec and BoW baselines. ChatGPT further shows robustness against noisy data, where the Word2Vec model achieves worse results due to noise. Results indicate that ChatGPT is a good generalist model that is capable of achieving good results across various problems without any specialized training; however, it is not as good as a specialized model for a downstream task.

Automated summary

ChatGPT demonstrates the potential of general artificial intelligence capabilities and performs well across various natural language processing tasks. This study evaluates ChatGPT's text classification abilities for affective computing problems including personality prediction, sentiment analysis, and suicide tendency detection. Results show that task-specific RoBERTa models generally outperform other baselines, while ChatGPT performs decently and is comparable to Word2Vec and BoW baselines. ChatGPT exhibits robustness against noisy data, outperforming Word2Vec in such scenarios. The study concludes that ChatGPT is a good generalist model but not as specialized as task-specific models for optimal performance.