GM Frens! Day 2 for the Fimio team at EthDenver was filled with all thing decentralized machine learning. We attended the countdown to the Filecoin virtual machine event I mentioned yesterday. It was a series of short talks catching everyone up to what has been happening in the filecoin network community. For folks who would like to watch the talks, they were livestreamed on EthGlobalTv, you can watch it here.
The thing that blew us away was the progress that has been made regarding compute over data, specifically where things stand with the Bacalhau project. Bacalhau is a platform for fast, cost-efficient, and secure computation by running jobs where the data is generated and stored. It basically allows you to do compute over data in a decentralized fashion.
This is a major breakthrough for anyone who runs large-scale machine learning systems like we do. Our long term goal at Fimio is to have a completely decentralized stack for machine learning. That means from where our data is stored, to the compute that runs our algorithms, to verification of the algorithm run, to decentralized inference. All along the pipeline we would like to have cryptographical verifiability. Our strength at Fimio lies in large-scale ML and not in design and building of decentralized infrastructure. So you can imagine how excited we were when we saw a demo of "Text to Image Auto-generating Stable Diffusion NFTS with Bacalhau." That demo let us know that Bacalhau has reached a state where it can handle large-scale ML.
Frens, I now I sound like a broken record, but this is a BIG DEAL!!! Why is this a big deal? That means that there are now certain data tasks that can be done where the entire data pipeline is secure and permissionless while doing compute. A prime example for the use case of such a system is electronic voting. We will unpack that use case in another blog. Till then, happy coding frens.
In case you missed it, you can see our Day 1 recap here.