The tool from the makers of ChatGPT still has a high error rate. In the future, a recognition tool should be able to distinguish between AI-generated and human-made content.
The company Open AI, which published the GPT chat, is now trying to get the consequences of this invention under control. Chat GPT is said to be able to imitate human language so well that many concerns have arisen about cheating on school and coursework or creating large-scale disinformation campaigns.
For this reason, a free tool called AI Text Classifier released. This should be able to distinguish between texts that were generated by artificial intelligence and those written by humans. The tool also indicates the probability that the text comes from an AI.
Made easy to use – but still unreliable
The use of the AI Text Classifier is deliberately chosen to be simple. You simply copy the desired text into the given text field and see the result immediately. Texts are then classified as “very unlikely”, “unlikely”, “unclear”, “possibly” or “probably written by an AI”.
However, the assessment of the new tool is still not very reliable. In test runs, the software - according to Open AI in a blog entry - correctly identified texts written by a computer in only 26% of cases. At the same time, nine percent of the texts formulated by humans were incorrectly assigned to a machine. The new AI's assessment also only works with a text length of 1000 characters or more, which entails a major limitation.
Furthermore, Open AI also points out that the tool was trained primarily on English texts. This means that the probability of errors is even higher with non-English texts.
For these reasons, it is recommended not to rely primarily on the assessment of the “classifier” when evaluating the texts.
Watermarks are also possible
The development company is also discussing the use of a type of digital watermark for Chat GPT. This would be designed so that it would not be visible to human eyes. However, special checking software would then signal with great certainty whether it is an AI text or not.
Great use in the future
As Futurezone reports, tools like the classifier are likely to be of great use in the future, even if their accuracy currently has some shortcomings.
For example, they could be used by professors to check their students' texts. This makes sense because, according to recent reports, students at Stanford are already using ChatGPT. This resulted in several final theses written by KI.
The Austrian Ministry of Education does not want to ban ChatGPT. Instead, teachers and students should deal with it critically. Maybe the AI Text Classifier can be of help.
Controversial chatbot – worry about the competition
The current hype surrounding Chat GPT is causing serious competition concerns. For example, the Google Group sees a threat to its own business model because, for example, search engines could no longer be used in such a wide range in the future. That's why Google is currently pausing the development of the software that can write and speak like a human and has been in planning for years. An internal email stated that a response to ChatGPT was a priority, meaning that they are currently trying to bring a bot onto the market that is very similar in use to ChatGPT.
Further information and reports on ChatGPT can be found HERE .
Sources:
Futurezone , The Standard
Also read our fact checks:
Molluscs: insects in food?
Misquotes from Bause and Özoğuz are circulating again
US government did not admit that 5G causes COVID-19!
Notes:
1) This content reflects the current state of affairs at the time of publication. The reproduction of individual images, screenshots, embeds or video sequences serves to discuss the topic. 2) Individual contributions were created through the use of machine assistance and were carefully checked by the Mimikama editorial team before publication. ( Reason )