AI bots are now outperforming humans in solving CAPTCHAs

Dec 17, 2023 ·

The study presented in the article “A Text-Generation Model Based on OpenAI GPT-3” explores how large language models like OpenAI GPT-3 can be used to generate meaningful text. The authors train a model based on GPT-3 on an open-domain dataset consisting of 1.2 million English articles and show that it is able to generate text with high semantic coherence. They measure the performance of the model using a variety of metrics including perplexity, BLEU scores, and human evaluation of the generated text. They also evaluate the model's ability to complete stories and generate answers to trivia questions.

The results show that the model is able to generate text with good accuracy and fluency. Specifically, it achieves a perplexity score of 8.4 (lower than the previous best result of 16.7), a BLEU score of 51.8, and an average human ranking score of 4.7 out of 5. Additionally, it successfully completes 92% of all story-completion tasks and answers questions with 83% accuracy.

Overall, this study demonstrates the potential of large language models such as OpenAI GPT-3 for generating meaningful text. By training the model on a large amount of data, it is able to generate text with high semantic coherence and accuracy. More research is needed to further refine these models and improve their performance even further.