Side note, but how defensible is Shutterstock's business after the generative AI boom?
I understand that Shutterstock's customers need royalty-free, high-res images with clear copyrights. Is the lack of clear copyrights the biggest hurdle to generative AI art adoption, or is the quality of the art itself the bigger hurdle?
The people writing articles that are... appropriate for AI generated mush didn't buy stock images, and organisations like e.g. the BBC are not going to put uncanny valley images of hospitalness on the banner of their articles.
Heise, one of the most reputable and in-depth tech publishers in Germany is using them. Mostly for articles where an image doesn't add much (like generic battery image in a text about new battery tech), but still. Possibly because images increase engagement
There are a lot of cases where a pub wants a graphic or image (perhaps depending on the pub's style), just about any graphic or image. Yes, for example, people are probably more likely to click on a social media link with an image.
Funny, because I live in the world of designers and a lot of them see a very near future where generated images replace 'real' photographed stock, if some are not there already.
It is already being used in commercial work. I know this from being in the industry, but I was also served an AI generated advertisement from a real company on YouTube today so it's in the wild too. It was a human talking about the product, you could tell it was AI by the way it reset it's face to the same position every sentence, and the stilted speaking. But it was honestly quite shocking, I could see someone not knowledgeable in tech being a bit weirded out but still convinced it was just a weird actor.
It will be in a lot of advertising really quickly, advertising uses a lot of image and footage fodder for lack of a nicer way to put it. "Lady walking with smile" slap the brand assets on it, hide the AI hands, done.
Stock images were also popular in advertising, since they were considered cheaper than hiring an illustrator or photographer. AI-generated images are cheaper again, and most people can’t tell the difference.
Former Shutterstock employee (was laid off in 12/2022).
Shutterstock makes most of it’s money from two operations:
1. Editorial content - images/video of news related events. Pictures of celebrities or photos of Biden, pictures of the Met Gala, etc. They have exclusive deals that keep this pipeline alive.
2. Commercial licensing - HBO/Netflix will often bulk buy a set of a couple hundred images for promoting their new show.
Everything else is trash web development and tech. They fired their AI team and their current AI tech is just rewrapped existing engines such as DALL-E. In the last five years they’re internal teams redesigned their website/marketplace at least 3 times. From React to React to NextJS. They frequently purchase companies and lay off the teams after about a year (see ya Giphy).
Shutterstock doesn’t have to be good at tech because they can feed off of the work of other companies doing it for them. That’s why they let companies like Facebook blatantly train off their data set. That’s why they consumed PicMonkey.
To make it seem like they are an innovator they rely on yearly announcements of investment in new fads (like their current AI efforts) and lay off tons of teams to boost their stock price so they can say they save on labor costs.
They are ripe to be displaced either by a smarter competitor that can offer better compensation per image or by Getty Images.
- Quality is still pretty big. And if I have to spend hours prompt engineering, is it really a win?
- That said, there is some copyright murkiness and I probably wouldn't use generative AI in a marketing campaign as a large company. A lot of more incidental use like conference presentations, I wouldn't worry too much.
For the near-term getting into the workflow is probably the key term. I'm hopeless with drawing but, for someone skilled with Photoshop and art generally, I can absolutely see getting generative AI output and then spending some time manipulating especially with all the new Photoshop ML tools.
I think they're well positioned. The winners of the generative AI space will be players with lots of copyrighted content. Shutterstock was already top of mind, and purchasing rights to all the images in Giphy strengthens their claim.
There's two paths. Either copyright claims lock down training, and Shutterstock now licenses their training dataset to companies; or generative art trained on open images lose their novelty (think Comic Sans but for photos) and Shutterstock gains an upper hand by having an existing pipeline of novel images to train with.
I understand that Shutterstock's customers need royalty-free, high-res images with clear copyrights. Is the lack of clear copyrights the biggest hurdle to generative AI art adoption, or is the quality of the art itself the bigger hurdle?