In the peak of AI era in early 2000s Photobucket became a leading provider of image hosting services. Which acted as the basis of popularity of platforms such as Myspace and Friendster, its user base increased to 70 million people to hold a large share of the US online photo market.
The current number of active users in Photobucket is just a million, down from its 70 million users years ago as reported by the analytic tracker Simliarweb. Although, there is a ray of hope, that platlers performance can be turned around in the wake of the AI revolution resurgence that could lead to new vitality of operations.
Ted Leonard, the CEO, who operates out of his base in Colorado, with a team of 40 employees, is in talks with different technology companies who want to sell the license for their software to Autonomous. This is as revealed in a report by Reuters. They are finding strategic alliances to partner with in order to license the extensive 13 billion images and videos of Photobucket. This purpose is to use the huge database for the training of generative AI models, which then is employed to create new content when the project receives textual commands.
He explained how he negotiated rates from 5 cent to $1 for photos and exceeded $1 for videos, some of which were even based on who is the buyer and what specific image the buyer desired. The big-name tech giants of the likes of Google, Meta, and Microsoft have been the main movers in the creation of generative AI models. At the start, OpenAI utilized vast volumes of freely available data which were scraped from the internet in order to train models such as ChatGPT which are in a position to emulate creativity. OpenAI has maintained the stance that such practice is both legal and honest but they have faced some challenges from the copyright holders who happen to object to the use of such information.
At the same time, such technology giants browse and cultivate connections that are covered in secrecy and watch series behind paywalls and login barriers. This result has sparked a black market of sorts that is home to all sorts of data coming from platforms that has seemingly gone extinct like chat logs and personal photos.
Technologically advanced companies insist that without any opportunity to mine web page data which is free-as-in(in terms of) a common pool administered by non-profit entity Common Crawl, developing such technology would be purely financially unfeasible. Nevertheless, their pattern has been plagued by copyright litigation and litigation, and the website owners have since responded by implementing technical blockers to prevent online scraping.
To meet this challenge, AI model manufacturers are utilizing methods to mitigate risks as well as to enhance their data supply chain security. This involves working with content creators to create partnerships and establishing relationships with a large number of data brokers to meet the need for data as the industry is growing.
One area that is coming into the light is the AI data firms holding the rights of a number of content sources including podcasts and videos. Defined.ai, a Seattle company, provides data for leading companies such as Google or Meta. Price varies from $1 to $2 and $2 to $4 for one image and short video respectively. Large films are $100 to $300 paid per hour. Instructional Images cost $5 and $7. Defined.ai put on ethics in data procurement, customers are given consent and data is anonymized. Companies pay their vendors 20% to 30% of the deal.
