Display the likelihood that a sample of text was generated by OpenAI's GPT-2 model.
OpenAI's recently released GPT-2 model has revealed itself to be capable of generating incredibly human-like text. With the rampant spread of fake news and reviews in today's world, such tools may pose a threat to the quality of the information found on the internet. Luckily, OpenAI also released a detector which is designed to detect whether a given portion of text has been generated by GPT-2 or not. This extension brings the detector to your browser via a simple browser extension. Simply select a portion of text (at least 50 words) and the extension will determine and display the probability of whether the selected text was generated by GPT-2 or not. Some notes: - The detector itself does not run in your browser. The extension makes HTTP requests to https://huggingface.co/openai-detector where the detector is publicly hosted and processes the response, displaying the results in your browser. - Results should be taken with a grain of salt. The detector is designed specifically to work with GPT-2 generated text. Using a different model or finetuning GPT-2 could be enough to avoid detection. In general, the detector is imperfect. Human-generated text may be marked as GPT-2 generated, and vice-versa.
- (2020-10-09) Giulio Starace: Hey, it does what it says it does! Would be cool to see it work with GPT-3 and generalise to other language transformers. PS the developer is very handsome
- (2020-01-23) HeyDuoDuosHeyDuoDuos HeyDuoDuosHeyDuoDuos: Good extension and useful, but sometimes fails to work.