To train the language models that form the basis of ChatGPT, such an inexorable amount of textual content is used today that the entire Wikipedia is only a small part of it. OpenAI sweeps the attic to dump the last forgotten forum into the well of the machine mind. However, South Korean researchers have found something they’ve never touched before: the dark web, the Internet’s underworld used by criminals.
The dark web is a network consisting of layers of encrypted pages and firewalls, which cannot be accessed using a regular browser, but only through the dedicated Tor application.
The artificial intelligence called DarkBERT has plunged into the world of digital black markets dealing with key vulnerabilities, stolen personal data, fake vaccination certificates, medicines, medicines and weapons with a noble intention: it is going after cybercrime.
DarkBERT is a bit like Google’s search bots called spiders, but they never venture into the wilds of the Tor network. The bot is technically based on the RoBERT architecture released in 2019. After the sudden success of GPTs, RoBERT has also been revived and they are doing more and more serious performance with it. The Korean team has built a huge illegal world database, and the machine has carved out the curriculum and is currently on
She knows herself on the field better than any other model.
According to the developers, the device is excellent for detecting newly emerging extortion programs or leaked information, and can effectively monitor the exchange of information in the depths of the forums.
The question is, what problems could arise from AI turning into cyberpolicing? We leave the possibility of responding to the US Navy and DARPA, which created Tor, as well as the CIA, who opened their offices on the dark web a few years ago and are still actively using it to communicate with their intelligence agents operating in Russia.
In any case, DarkBERT is not open to the public and is currently only available to academics. It’s a serious milestone in dark web research and shows its presence in the future as language paradigms are separated into application domains and niche domains.
(futuristicAnd IFL ScienceAnd The origins of innovation)
