In February 2024, Reddit with Google concluded an agreement of $ 60 million to use the search giant data on the platform to train its artificial intelligence models. In the discussions were mainly the missing Reddit users whose data was sold.
The agreement reflected the Reity of Modern Internet: Large technology companies own virtually all our online data and decide what to do with this data. It is not surprising that many platforms will monetize their data and a rapidly growing way to achieve it today to sell companies that are in themselves massive technology companies using data to train increasingly strong models.
The decentralized bathtub platform, which began as a class project for MIT, is on the mission to return the users power. The company has created a fully user network that allows individuals to record their data and control how they are used. AI developers can build users on new models, and if users agree to contribute their data for training, they will acquire proportional ownership in models.
The aim is to provide each share in AI systems that will increasingly shape our company and at the same time unlock new data funds for progress in technology.
“These data are necessary to create a better AI system,” says the co -founder of the bathtub Anna Kazuauskas ’19. “We have created a decentralized system to get better data – which now sit inside large technology companies – while allowing users to maintain final ownership.”
From economy to blockchain
Many high school students have pictures of pop star or athletes on the bedroom walls. Kazlauskas had a picture of US Finance Minister Janet Yellen.
Kazlauskas came to MIT to become an economist, but ended up as one of the five students who joined the MIT Bitcoin Club in 2015, and the experienced led her to the world of blockchains and cryptocurrencies.
From her dormitory in MacGregor House, the Ethereum cryptocurrency began to benefit. She even occasionally explored the campus dumps in finding discarded computer chips.
“I am interested in everything around computer science and networking,” says Kazlauskas. “This included distributed systems from the point of view of blockchain and how they can move economic power to individuals, as well as artificial intelligence and economic.”
Kazlauskas builds Art Abal, who then waited for Harvard University, in the class of Media Lab Emmergent Ventures, and the couple decided to work on new ways to get data for training.
“The question was: How could you have a large number of people who contribute to this AI system using a more distributed network?” Kazlauskas remembers.
Kazlauskas and Abal tried to solve the status quo, where most models are trained by scratching public data on the Internet. Large technology companies often also buy large data sets from other companies.
The founder’s approach has evolved over the years and has been informed about Kazlousk’s experience, who worked at the financial blockchain cell after graduation. Kazlauskas, however, attributes its time to MIT, helping her think about this problem, and an instructor for Emergent Ventures, Rameh Raskar, still helps the bathtub to think about AI research issues today.
“It was great to have an open father -in -law that just builds, hackled and explored,” says Kazulauskas. “I think ethos on mit is really important. It’s just building things, seeing what works, and continue the iteration.”
Today, the bathtub uses little -known law, which allows users of most large technological platforms to directly export its data. Users can upload this information to encrypted digital wallets in the bathtub and pay off to train models as they see fit.
AI engineers can propose ideas for new models with an open source code and people can associate their data to help train the model. In the world of blockchain, data funds are called DAOS data, which means a decentralized autonomous organization. Data can also be used to create personalized AI models and agents.
In the bath, data is used in a way that represents user privacy because the system of unidentifiable information. Once the model is created, Hauttion users ownership so that every time they are used, they are extended rewarded based on how much their data they have helped to train.
“From the developer’s point of view, you can now create these hyper personalized health applications that take into account exactly what you ate, how you slept, how you practice,” says Kazlauskas. “These are not applications today, because of these brick gardens of large technology companies.”
Crowdsourced, user -owned AI
Last year, the engineer of machine learning suggests using user data of the bathtub for training AI, which could generate Reddit posts. Their Reddit data has contributed more than 140,000 bathtub users that contained posts, comments, messages and more. Users have decided on the conditions in which the model can be used, and after it was created, they santinated the ownership of the model.
The bathtub made it allowed similar initiatives with users with design data from the X -Social Media platform; Sleep data from sources such as oura rings; And others. There are also collaborations that combine data funds and create wide AI applications.
“Let’s say users have Spotify data, Reddit data and fashion data,” Kazlauskas explains. “Ussully, Spotify will not cooperate with the companies of companies, and in fact there is regulation against it.
The bathtub has over 1 million users and more than 20 live DAO data. Users in the VANA system offer more than 300 additional data, and Kazlauskas says that many of them will go into production this year.
“I think there are many promised in generalized AI models, personalized medicine and new consumer applications because it is difficult to combine all this data or gain access to the first place,” says Kazlauskas.
Data funds allow users to achieve something that even the strongest technology companies are fighting with today.
“Today, large technology companies have created these data ditches, so the best data sets are not available to anyone,” says Kazuauskas. “It is a problem with a collective event where my data itself is not so valuable, but the data fund with tens of thousands or millions of people is really valuable. The bathtub allows you to build these funds. It is mutually advantageous: users benefit from AI, because models.
(Tagstotranslate) bath Network (T) Anna Kazlauskas