Should we preserve pre-ai internet before it contaminates?

Ai Chatbots Arrival marks a historical line of dividing after online material is not fully reliable to be human, but how can people turn this change? While some quickly work in the archive “undefined” data from the pre-ai era, AI releases themselves we need to record how the future historians are going to improve.

Rajiv pantinga businessman and former official of the principal technology of both The New York Times and The Wall Street Journalsays he saw AI as a risk of information such as news stories that turned out to be part of the historical record. “I think about this ‘digital archeology’ problem since chatgtt has been launched, and becomes more urgent every month,” says Pant. “Today, there is no reliable way to identify the person’s content written from AI’s material. This is an academic problem to scientific acting.”

consideration John Graham-Cumming In the cybersecurity firm floudflade, the information held before the end of 2022, when chatgpt is launched, same as low-background steel. This metal, simelted before the Trinidad nuclear Bombs Tests on 16 July 1945, valued by the bright contamination from the atomic weats era making noise in readings.

Graham-cumming makes a website called lowbackspoel.i In archives of origin of data not contaminated with AI, such as a perfect download of Wikipedia From August 2022. Studies have already shown Wikipedia today Showed signs of great AI input.

“There’s a point where we do everything on our own, and after a few times we start to raise these chat systems,” he said. “So the idea is to say – you can see this contamination, or you can do it a kind of a vault – you know, people, we come here. And we’re here. And we’ve got this point, we get more help.”

Mark Graham Wayback machine is running Internet archive, an internet internet-archive project in 1996, says archive stores in 160 terabytes of new information every day.

Instead of preserving pre-ai internet, Graham wants to start creating AI output archives for future researchers and historians. He has a plan to start asking 1000 topical questions a day in chatbots and keep their answers. And because it’s a big task, he’ll use AI to do it: AI that records ai’s output change, for the curiosity of future people.

“You ask for it a specific question and then you can get an answer,” Graham said. “And tomorrow you ask it with the same question and maybe you can get a little different answer.”

Graham-cumming quickly focuses he is not anti-ai, and that preserving the man-made manufactured in AI models. That is because the low quality output of the AI repositioned in new models can have a harmful effect, leading what it is known to be “Putting the model“. The avoidance of this is a useful effort, he said.

“At some point, one of these AIS is thinking of something we don’t think of people. Prove that something math,” as I have something to be new.