Don’t want ChatGPT to suck knowledge from your site? It is now possible!

After many months of controversy over the collection of data to train Generative AIsOpenAI is paving the way for a game-changer.

Does the future with artificial intelligence suddenly look a little more serene?

A small text file to edit

ChatGPT’s database is a bit dated, being limited to 2021. This is a problem for its developer, OpenAI, which faces increasing competition, especially from Google Bard, which has access much more direct to the entire Web, and therefore potentially to its knowledge. To overcome this deficiency, a certain GPTBot has just appeared. Just like search engine crawlers that crawl the web to reference website pages, it has the ability to wander from page to page, suck up data and pass it on to ChatGPT.

A vision of horror, you say? Not quite, because OpenAI has thought of everything. Indeed, website administrators can now prevent GPTBot from exploring their content by blocking their IP address, or by modifying a file they already know very well: robot.txt. The latter acts as a bouncer in front of a site, authorizing or not the passage to robots from Google or Bing, to name a few. That of ChatGPT is therefore no exception, and this is very good news, which could mark the beginning of major changes for the sector in the best direction.

A significant precedent for an entire sector?

Between writers’ strikes, artist complaints and the immediate need to create new regulations, programs such as ChatGPT or Midjourney have caused quite a stir, to say the least. Needing to collect a lot of knowledge to train themselves, they naturally turned to the Internet to get as much as possible. However, the protection of the personal data of Internet users and the intellectual property of creators and other organizations has quickly become a sensitive point, to the point of causing numerous outcry.

Artificial intelligence has the potential to bring enormous changes to our habits, for good or ill. However, their use raises questions that our societies have not yet succeeded in fully addressing and on which there is no consensus. The generative AIs developed by the tech giants will therefore have to adapt a little, otherwise they will receive a huge backlash which could well slow down their development more than anything else.

This is why the sector works hand in hand with government institutions to adapt serenely to our societies. Evidenced by an agreement signed with the White House by several AI companies, with the aim of developing a system to know if something has been generated by artificial intelligence or not. And if not everyone promises to stop using user data to develop their products, it could be that many critics of ChatGPT and company find OpenAI’s approach inspiring. However, the latter does not refrain from specifying that ” Allowing GPTBot to access your site can help AI models become more accurate and improve their overall capabilities and security “.

Source : The Verge