Fusing AI & Web3: Improving the world one dataset at the time
Since 2022, Datafund has actively expanded its footprint in the artificial intelligence (AI) domain. This strategic move was driven by the clear understanding that AI adoption is not optional; businesses that fail to leverage AI risk obsolescence.
Building independence in AI technology
We decided to tackle this transition by building our AI infrastructure from the ground up. While it’s true that the largest AI models require an insane amount of compute, smaller models are also becoming progressively efficient and capable. And with the option to throw different agents at a task simultaneously, the range of what can be done, even without a multi-billion data centre, expands dramatically.
Not only that, owning our hardware has given us all the freedom we need to explore, learn and build with AI-related technologies.
Reshaping medical diagnosis
The decision to start treading the AI waters soon proved to be the right one. In 2022 we forged a partnership with Slovenia’s biggest medical institution, the University Medical Centre Ljubljana (UKCL), to carry out a pilot project in the field of haematological cytology. By combining anonymised, real-life images of patients’ bone marrow smears with AI training models, we have created our own prediction model.
It will help medical professionals identify types of cells inside the bone marrow much faster. This will lead to a better, quicker diagnosis and treatment, while relieving medical personnel of having to go through thousands of images by hand.
UKCL is not the only one who recognised the importance of this project and offered us to expand our collaboration to other medical fields. Our pioneering work also earned us a spot in the prestigious Nvidia Inception accelerator.
A setup for the future
The ongoing work we’re doing with UKCL is proving to be a great diving board for our future endeavours. It’s allowing us to build proprietary, high-quality datasets that can be used for training Datafund’s AI models.
Creating such models perfectly aligns with our goal to create even more advanced tools for predicting medical events. Because of the flexibility that they offer the medical field is only one possible avenue of exploration. We’re certain that we’ll be be able to add new use cases in the not-so-distant future.
Providing data provenance with Web3
The matter of creating proprietary models is tightly tied into the question of the kind of datasets we’ll be using to train them. In a sensitive field such as medicine, the quality and origin (provenance) of datasets matters, especially in the light of the new EU AI Act.
That’s why we have already built a fully functional tool that can attest to a dataset’s provenance. It’s fully Web3-embedded, so the whole attestation and notarisation process unfolds in a private, secure and decentralised manner using blockchain and Swarm’s distributed storage. In addition, fusing AI with Web3 immediately opens up the possibility to monetise models, datasets or tasks.
All of the above positions us at the front edge of the curve. We have the knowledge and experience to embrace the AI shake-up heading our way. The future is here and we’re excited to take it head on!
