The development of AI

The development of AI  

The EU rulebook for AI is taking shape – requires transparency  

An upcoming EU regulation of AI tools is starting to take shape, writes Reuters.  

According to a first draft from the EU Parliament, services and tools are to be classified based on the risks they may pose, from "minimal risk" to "unacceptable risk", according to Reuters.  

High risk tools will not be banned. Instead, there must be high demands on transparency.  

Companies that develop generative AI such as the chatbot ChatGPT or the image tool Midjourney must also be forced to state which copyrighted material the AI was trained on. 

Google's AI Bard/Nordic resistance movement. TT  

The development of AI  

Google trains its AI on a Swedish Nazi website  

Google's C4 dataset lets its artificial intelligence train on the Swedish Nazi Nordic Resistance Movement's site Nordfront, reports SVT. The TV channel has taken part in analyzes carried out by the research institute "Allen Institute for AI" together with the Washington Post.  

SVT and SR are also on the list. So does the left-wing extremist network Antifascistisk Aktions' website, the Radiation Protection Foundation, which spreads conspiracy theories about 5G, and vaken.se, which spreads anti-vaccine content.  

According to Thomas Nygren, professor of didactics at Uppsala University, Nordfront in particular - which is such a small site - will not have any direct effect.  

- But similar data texts from similar sites have had a direct meaning. This has been seen by asking the AI about, for example, the Holocaust or about homosexuals, he tells SVT Nyheter.  

Google states that it "works hard to minimize potential risks to users". 

"We have developed a number of different methods and technologies to prevent bad results when using these models," Andrea Lewis Åkerman, communications manager at Google Sweden, writes to SVT Nyheter.