AI and DePIN — Solana Spaces | ETH Denver [BTN: March 1st, 2024]
Over the past two weeks, the Render Network has participated in a number of conversations at the forefront of decentralized AI. Meanwhile, digital art events like NFT Paris have featured artists from across the Render community and Beeple kicked off his Digital Art Death Match with a call for submissions. Let’s dive in to the latest from the Render Network.
Web3, DePIN and AI Spaces with the Solana Foundation
Last week, the Render Network’s Jules Urbach and Trevor Harries-Jones joined Co-founder of Solana Anatoly Yakovenko alongside Kuleen Nimkar and Dare Sobande from the Solana Foundation in a Space exploring the convergence between AI and web3.
The Space featured leaders at the forefront of decentralized AI, such as IO.net, GRASS, and Synesis One, and explored a wide range of themes — from decentralized compute infrastructure and cryptography to digital rights, peer-to-peer generative AI, and Solana DePIN.
The conversation started by exploring how blockchain infrastructure can play a critical part in different stages of AI development — from data collection and labeling to model training and inference. Panelists discussed the advantages of decentralized approaches to AI, which included reducing censorship risks, increasing accessibility, generating greater scale, as well as new tools for data ownership and more efficient resource coordination.
As Jules described, the combination of decentralized scale and immutable public data that DePIN offers can be as transformative for AI computing as it has been for 3D rendering and GPU graphics applications.
“In the case of one layer to what we are providing, DePIN…studio movies have been rendered on decentralized nodes in our system with end-to-end encryption — you’re getting faster speed, you’re getting one tenth of cost, you have massive availability, and a ledger can authenticate what an image is or where the data came from, that’s built into the crypto part of it — we have Proof of Render.
Generative AI on the inference and training side also fit to the kinds of GPUs that are already on Render. You don’t need an H100 to do training on generative video. One of the more exciting propositions for us is we look at the Apple ecosystem where you have GPUs with 120 gigabytes of video memory, those are out there in the tens of millions. Tapping into that is going to be a massive win over the cost of the scaling issues we have on centralized clouds. That’s where I see huge advantages of decentralized approaches like we have.” — Jules Urbach
Later on during the space, Trevor discussed these elements in the community’s decision to transition the Render Network to the Solana blockchain, enabling it to leverage infrastructure with lowers costs, increased speed, and new toolsets for on-chain provenance.
“Last year we had a spirited community vote on this topic and I think the key driving factors in our blockchain selection were cost and speed…And then lastly, having a mature DePIN community in the Solana ecosystem really helped us.
We believe proof of creation is gonna become more and more of a central issue in the age of AI. For us, compressed NFTs were really interesting as a key on-chain part of the solution for this. So technology was really central.”
— Trevor Harries-Jones
The conversation moved on to the topic of authentication and security as AI expands into more sectors like media and entertainment, raising critical questions about IP monetization and digital rights. One of the key challenges is how to properly verify what data or IP is used in training across AI image or video generators. Jules discussed how Proof of Render could be an important building block for digital rights in an age of AI, where the integration of decentralized GPU computing and blockchain ledgers provides a proof of work model and data layer for AI applications.
“Having the ability to take your identity visually or your voice and tag that and put any sort of rights and restrictions you want around that is exactly why we wanted to build Render on a blockchain and where we look to leverage Solana in the future. Everything that goes into a training or inference step, at least on Render, we can tag, hash, and provide a receipt for, and even payout royalties right to contributors. It goes back to the value proposition of an NFT, which in my mind shouldn’t be the JPEG, but the actual data and 3D scene.
It can all be built into a smart contract that pays things out, similar to people getting royalty checks for being in a TV show that go into syndication. That’s the future where I see the value proposition of the blockchain and verifiable generative AI on systems like Render really changing the game.”
— Jules Urbach
Jules concluded by discussing innovations like custom ASIC capabilities built into newer Apple GPUs which can run AI work using lower energy, enabling training and inference on a wide range of general purposed GPUs. He also noted the overlap between some of the GPU processes in rendering, like GPU raytracing, and the work that is done in text-to-image diffusion models — a conversation picked up on X with the Stable Diffusion team, exploring the convergence of 3D and AI computing in DePIN.
Render Network joins R23aI World at ETH Denver
Earlier this week, the Render Network’s Trevor Harries-Jones joined a panel discussion with leading DePIN experts from Geodnet, WifiMap, and Grass at R3al World at ETHDenver.
The session, led by Connor Lovely the lead for DePIN at IoTeX, explored new DePIN applications with a focus on community building, adoption strategies, and bluesky approaches where web3 networks and AI converge to build new types of decentralized utilities.
The discussion delved into advancements in deep learning technology, along with strategies for business development to bootstrap DePIN networks. Trevor explored how the Render Network has expanded to AI in addition to its existing use for decentralized 3D rendering. He explained that high-end GPU power (like H100 models used in Data Centers) aren’t always necessary for tasks such as inference, diffusion and text-to-video processing, noting that many developers prefer using less powerful, consumer-grade GPUs at greater scale (over 250 million addressable units). He highlighted the massive opportunities for leveraging the scale of latent consumer compute for AI applications outside of LLM training that don’t require the highest performance GPUs.
Trevor concluded the AI question by sharing his excitement about the growth in new diffusion based models like Sora and how the Render Network’s infrastructure can be used in applications like diffusion video, complementing the network’s use for decentralized GPU rendering.
We’re really excited about Diffusion models like Sora from just two weeks ago. What we see coming this next year in terms of diffusion models and other models besides LLM’s is demand really coming from those scenes and finding the right product market fit.”
Of note were the benefits of having multiple applications on the network from 3D rendering to AI general purpose computing. Because the market pricing for 3D rendering generates roughly 8x revenue per job compared to general computing jobs, multiple application streams help build network effects, where a variety of workflows become complementary in generating more consistent supply and demand. For example, the aggregation of diverse workstreams enabling the network to leverage more latent consumer GPUs by improving the economic efficiency of the network.
NFT Paris 2024
NFT Paris 2024 touched down last week at the Museé Grand Palais in Paris featuring a number of Render Network creators alongside artists, creatives, and executives from some of the biggest brands in the world, like soccer juggernauts PSG and FC Barcelona, and fashion brands Bulgari and Louis Vuitton, and leading contemporary digital artists.
One of the larger showcases during the weekend was Superchief Gallery’s “Mapping the New Normals” exhibit, which features familiar names with Render Network luminaries like David Ariew, Justin Bodnar, Josh Pierce, (JSTNGRAPHICS) and Nicole Ruggiero amongst the lineup.
Finally, this week — in the aftermath of NFT Paris — the Bloom collective is showcasing the work of Render Network user Hannes Hummel in their gallery event“Lux.” The event will close on March 2nd, so we suggest seeing this amazing art while it’s available, live and in-person.
Beeple’s Digtal Art Death Match — Powered by the Render Network
This week Beeple sent out a call for submissions for a Digital Art Death Match on May 4th — a live immersive event with real-time digital art contensts, talks and artist showcases, sponsored by the Render Network and featuring many amazing artists from the Render community.
Artists can submit their submissions and the public can sign up for the waitlist here.
Join us in the Rendering Revolution at:
Website: https://render.x.io
Twitter: https://twitter.com/rendernetwork
Knowledge Base: https://know.rendernetwork.com/
Discord: https://discord.gg/rendernetwork
Telegram: https://t.me/rendernetwork
Render Network Foundation: https://renderfoundation.com/
