2/ The experiments involved deploying a CNN trained on MNIST to each protocol and evaluating the live solution
looking at: decentralized hosting, hardware support, inference job types, model management, economics, and privacy
thanks to @AlgoveraAI for supporting this work
4/ AI & Computing:
Filecoin’s role in AI innovation is significant. Teams like @SingularityNET , @MarlinProtocol , and @AlgoveraAI are using Filecoin to power compute resources and storage solutions for AI projects, unlocking the potential of decentralized AI. 🤖💡
The year is 1610 and the American Frontier has just been opened. Do you build your new company/community (a) under the feudal system in Europe or (b) with sovereignty in the New World?
The year is 2024. Do you build your AI project (a) under SB-1047 (and future AI regulation) in SF or (b) with sovereignty on decentralized infrastructure?
Cool demo! True multi-agent workflows *require* decentralized infrastructure and Web3. Excited to see all of our R&D on decentralized storage, compute, and marketplaces over the past 3 years being put to good use 👏
1/ What people call multi-agent systems are really single agent, since the same LLM & codebase are used. Here's Yohei (creator of BabyAGI) making that case in a recent thread. https://t.co/ozLFz5Y5pm
Great news! Our PR to @ocolegro's https://t.co/LtdkQMNWkt was just merged https://t.co/GvUqLB6LL9 You can now use @ollama to generate synthetic datasets locally (with RAG).
Next up: Running synthetic data generation workflows across our decentralized network of compute nodes 🚀
@LizaTitmus31745 Cool! So maybe something like (i) an assistant with knowledge about the impacts of plastic bottle use, or (ii) some AI-generated images/videos to bring attention to the subject?
1/ Is there a task that you would like to solve with AI, that you don't currently know how to solve? Maybe you're chatting with ChatGPT about it, but getting stuck? Or maybe you want a personalized assistant for some specific use case but don't know how to build it?
3/ Feel free to DM us, post in the comments, or in the the #🤔│post-a-task forum on our discord (https://t.co/tcBuCg2gm0), and we'll share with our network of AI builder communities that can work on it for you.
2/ It could be something easy like "Create a tweet thread for a research paper" or "Create a cool image or video of X", or medium-level like "Create a chat bot for answering questions about open source LLMs" or as hard as "Create a website that achieves Y".
This was an awesome event organised by @NousResearch @Ollama_ai et al. We can't wait to build more open, local and decentralized AI infrastructure and tooling with this growing community.
excited to see @theemozilla officially launch @NousResearch and Nous Forge today!
with many many open source AI luminaries including @teknium@winglian@jeremyphoward et al in attendance. Feels like the start of something big.
Recently @ValoryAG awarded a grant to Algovera to explore the intersection of Crypto x LLM agent frameworks. We created LLM building blocks (skills, short-term/long-term memory) in their framework & two decentralized chat/question answering assistant apps. https://t.co/kj9gqcUIKX
Some cool AI research (sponsored by Algovera microGrants) that explores how using recurrence with attention in transformers can help to extrapolate to unseen datapoints during inference/deployment! Check out the original proposal here https://t.co/nvlPv3t6JH
Under what priors and inductive biases can attention-based models OOD extrapolate?
I did a short investigation into Adaptive computation mechanisms. What I found - recurrence is a strong inductive bias for OOD length-extrapolation... 🧵
Each of the last 3 of these webinars has been jam-packed with really interesting ideas. If you're working on anything related to decentalized AI, consider applying to speak!
Working at the intersection of Crypto and AI? Up for sharing a 10-minute talk with a passionate audience on September 7th?
The next Crypto x AI Mini-Conference is accepting applications!
Selected speakers get to present what they're most excited about, take questions, and more.