MicrostockGroup

Agency Based Discussion => Shutterstock.com => Topic started by: Jo Ann Snover on December 14, 2024, 20:02

Title: Shutterstock "research license" for data training
Post by: Jo Ann Snover on December 14, 2024, 20:02
Apparently announced on Friday:

https://venturebeat.com/ai/shutterstock-pioneers-research-license-model-with-lightricks-lowering-barriers-to-ai-training-data/ (https://venturebeat.com/ai/shutterstock-pioneers-research-license-model-with-lightricks-lowering-barriers-to-ai-training-data/)

"The new licensing model addresses a critical challenge in AI development: the high cost of accessing quality training data. It enables companies to start with a smaller research license for testing and experimentation before committing to more expensive commercial licenses."

Later on the article says contributors get 20% of the data licensing revenue.

https://www.gurufocus.com/news/2632851/shutterstock-inc-sstk-partners-with-lightricks-for-ai-video-model-training (https://www.gurufocus.com/news/2632851/shutterstock-inc-sstk-partners-with-lightricks-for-ai-video-model-training)

https://www.stocktitan.net/news/SSTK/lightricks-partners-with-shutterstock-for-video-training-data-to-55famkrcfdtz.html (https://www.stocktitan.net/news/SSTK/lightricks-partners-with-shutterstock-for-video-training-data-to-55famkrcfdtz.html)

In the last article above it refers to SS's "billion-plus" asset library. Not sure where that comes from - the web site says over 530 million assets
Title: Re: Shutterstock "research license" for data training
Post by: cobalt on December 14, 2024, 20:17
Probably includes the declined comtent.
Title: Re: Shutterstock "research license" for data training
Post by: gnirtS on December 15, 2024, 06:33
In the last article above it refers to SS's "billion-plus" asset library. Not sure where that comes from - the web site says over 530 million assets

Likely all the stuff rejected for data licensing.  Maybe also assets of people they've bought like Bigstock/P5 etc etc
Title: Re: Shutterstock "research license" for data training
Post by: bennu99 on December 15, 2024, 07:10
Not to sound cynical, but I’d be really curious to see what they actually achieve with this "high-quality video library." If they’re including Envato’s video assets as part of their dataset, they should be transparent about it. As far as I remember, Envato allowed users to incorporate AI into their work, as long as the final product wasn’t just a direct output from AI. That policy may have changed (I haven’t kept up since I never used it that way), but I know there are plenty of assets in Envato’s library that might not disclose whether AI played a role, especially during the time it was permitted.
Title: Re: Shutterstock "research license" for data training
Post by: stocker2011 on December 15, 2024, 07:54
In the last article above it refers to SS's "billion-plus" asset library. Not sure where that comes from - the web site says over 530 million assets

Likely all the stuff rejected for data licensing.  Maybe also assets of people they've bought like Bigstock/P5 etc etc

More likely all their videos not just the rejected, and as you mentioned, assets from all the companies they bought.

"Contributors can also opt out of having their content used for AI training, though Mandell notes only about 1% have chosen to do so."

I suspect only 1% have 'chosen' to do so because the rest didn't know how or where to opt out.