Hacker News new | past | comments | ask | show | jobs | submit login
Japan will not enforce copyright laws on AI data training, says minister (diyphotography.net)
17 points by mikece on June 6, 2023 | hide | past | favorite | 7 comments



This is a rehash of a previous article, which is a complete lie. I summarized what the original Japanese source says a few days back:

https://news.ycombinator.com/item?id=36147226

Slides published by the Japanese government shows that they're evaluating how to deal with copyright and AI. Their view is that current laws has few (but not zero) restrictions when it comes to training AI. However, strict copyright restrictions do apply to AI generated works even today. If it's proven that an AI-generated work is similar to or based on existing copyrighted material, that would be regarded as a violation and criminal charges may follow.

Here's the slide (in Japanese):

https://www8.cao.go.jp/cstp/ai/ai_team/3kai/shiryo.pdf


It's disappointing to see reporting based on machine translations. Unfortunately it's not at all uncommon and many people will accept the gibberish mtl outputs at face value.


Interesting, given how stringent Japanese copyright is otherwise. I'd bet large Japanese media companies lobbied against that (eg Nintendo surely doesn't want AI nets able to make Mario/Zelda/etc art indistinguishable from source material). Curious what ultimately led them to reach that conclusion.


Partly due to the stringent Japanese copyright laws that hindered their progress in technology development, such as search engines, they found themselves falling behind and allowing outsiders to gain control. Undoubtedly, they wanted to avoid repeating that mistake, which likely explains their current situation.


My guess is this is to drum up support for even stricter laws. "Nope nothing we can do to stop this with current laws" etc


This is the worst (or best) version of Prisoners Dilemma

All it takes is one G20 country to allow this to happen. And because of the potential upside of having a homegrown AI industry, it's almost guaranteed that someone will choose to betray.


I remember when three Silicon Valley companies were going to announce a new standard Unix. The announcement day came, all three announced their deviations from the standard to make their version best.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: