zlacker

Access to ressources is hardly a new problem: when I was an NLP graduate student about a decade ago a teacher of us had scrapped (and continued to do so) a major newspaper for years to make a corpus. The legality of that was questionable at best, yet it was used in academic paper and a subset for training.

The same is equally applicable to image: Google got rich in part by making illegal copies of whatever image he could find. Existing regulations could be updated to include ML model but that won't stop bad or big enough actors to do what they want.

> We’re in a “mutually assured destruction” situation now

No, we aren't. Very good spam generators aren't comparable to mass destruction weapons.