Human vs AI: Evaluating the Effectiveness of Deep Learning in AI Text Detection

This study looks on the effectiveness of fine-tuned DistilBERT and LSTM models for distinguishing between human-written and AI-generated text. We address the growing demand for reliable AI text identification as an outcome of the potential misuse of large language models (LLMs) in instances related to the transmission of misleading data. This approach combines advantage of the beneficial characteristics of both model, LSTMs’ capacity to extract sequential information from text input, and DistilBERT’s pre-trained interpretation of contextualize associations. We build these models employing a set of labels that includes writings created through machines and people indiscriminately. The evaluation metrics exhibit outstanding achievement, with DistilBERT and LSTM achieving 100% classification accuracy (1.00) of the test sample. By stimulating the ethical use of LLMs while minimizing any adverse effects, the study advances the development of dependable methods that can identify text composed with machines with intelligence.