Just finished a deep-dive on the wonderful LogUp, an elegant GKR-based lookup from @PapiniShahar & @UHaboeck! Thanks everyone for the discussions and suggestions so far. https://t.co/UpWwTA7MHG
Exciting times for the ZKML space as Zero Gravity continues its work exploring and developing the use of zero-knowledge proofs for Weightless Neural Networks (WNNs) .
Read the blog post by @elus_aegis and @georgwiese to learn more!
https://t.co/HJ1tQPMmuq
Here is a write up of some recent discussions we had about "deep differentiable logic gate nets" and their potential future relevance to #zkml
https://t.co/cW4ND3LJuq
Join us May 2 @UCBerkeley or online for ZKP Workshop, Hackathon Demos & Awards Ceremony 🎓
With ~4K ZKP MOOC participants, ~600 hackathon participants & ~150 teams from over 60 countries, ~30 finalists competing for $200K+ in prizes! 🏆
Register now 👉 https://t.co/v8Tfq16kXv
Will there be a day when we elect LLMs as our democratic representatives, and use ZKML to prove that it is indeed the elected model making decisions on our behalf? Or just pit them against one another to simulate parliamentary discussion, then make the decisions ourselves?
@recipromancer@0xJeeva@tahpot Jokes aside, I am personally _very_ excited about getting these models to talk to one another, even two instances of the same one. What a resource!
1st 🏆: @drCathieSo_eth, @SecuritaAegis, @georgwiese, @benwilson_ml, @VicSintNic, @unzvfu - ZeroGravity, a weightless neural network architecture to reduce the proving of inference to lookup arguments on a highly optimised Bloom filter implementation.
https://t.co/13J0hveQKw
Awesome Zero Gravity winning team at @__zkhack__ this weekend took an #ml model without floating point arithmetic and implemented it for #zkps without loss of accuracy.
The "weight" is over!
🏆 We won 1st place at @__zkhack__ Lisbon with "Zero Gravity", which marries weightless neural networks & zero-knowledge proofs for trustless ML inference without performance tradeoff. Explore our work: https://t.co/g6P2i0Ayzj
@VicSintNic @SecuritaAegis
Calling a smart contract an "autonomous object" seems a prefect misnomer. They can't _initiate_ action, only act at the call of off-chain agents i.e. humans or bots. I want truly autonomous smart contracts, with initiative & agency. Any ideas? Does this exist? #Ethereum
A smart contract could incentivise off-chain agents to call it every block. Seems a hack though. Shouldn't they be able to bid themselves on being called again in the next block, i.e. by the EVM itself?
I just read the MEV paper https://t.co/UtQaJTxmwC (so late to the party!). A great example of how smart contracts can transactionally bundle multiple message calls s.t. they are executed in an all-or-nothing manner. Simple but powerful!
@matthiasroder Perhaps it's been hard the few years to get good GPUs, including for GCP and AWS? On both of them you can apply to have the quotas raised, but they don't raise them very far ...! There is also https://t.co/GOpXXSyEih but I haven't tried it yet