🌵 The Intersection of Crypto & AI 🌵

Big Brain Breakdown

Market Metrics

Total Crypto Market Cap: down 0.9% to $2.72T
Total AI Sector Market Cap: down 2.9% to $34.8B

Top Movers (24hrs):

📈Spectre AI (SPECTRE): up 25% to $3.07
📈Roko Network (ROKO): up 20.3% to $0.00005742
📈DeepFakeAI (FAKEAI): up 19.1% to $0.01129

Daily News

🟠 Nvidia's Q1 earnings beat expectations, with the company announcing a 10-for-1 stock split, increased dividend, and strong demand for generative AI driving data center growth, causing its stock to rise nearly 7% and trade above $1,000 in pre-market trading.

🟠 NetMind has captured significant mindshare after launching a usage-based mining mechanism that rewards users based on their machines' actual usage within the ecosystem. The team also shipped a GPU dashboard that provides comprehensive insights into GPU counts, utilization rates, and overall network contributions.

🟠 AIOZ Network has become the first DePIN project listed on the Nvidia Accelerated Applications Catalog, allowing Nvidia's global user base to explore AIOZ Network's capabilities for application development, including web3 AI computes, data storage, and streaming infrastructure.

🟠 NEAR Protocol has launched NEAR AI, a research and development lab focused on decentralized AI, along with an AI-focused incubator and investment wing to fund and support AI projects built on the NEAR ecosystem.

🟠 The SEC is set to make its decision today on the approval of a spot Ethereum ETFs based on recent court rulings and its past decision to allow ETH futures ETFs. According to prediction markets and thought leaders, there is a high likelihood that the ETF is approved.

🧠 Big Brain Breakdown

Welcome back to another Big Brain Breakdown, where we help you understand the fundamentals of blockchain AI projects so you can stay ahead of the herd and invest in projects poised for outperformance. Today we are breaking down decentralized inferencing, the process of using a trained model to make predictions or decisions based on new, unseen data. It's the stage where the model applies its learned knowledge to real-world scenarios.

Think of inferencing like a student applying what they've learned in math class to solve new problems. During the learning process, the student is given a set of math problems along with their solutions. They study these examples, understand the patterns, and learn how to solve similar problems. Once the student has gained the necessary knowledge and skills, they are ready to tackle new math problems on their own.

Similarly, in the inferencing phase, a trained AI model is presented with new, unseen data and is expected to provide accurate predictions or decisions based on what it has learned during training. The model analyzes the input data, applies the learned patterns and representations, and generates an output, such as a predicted class label or a numerical value.

Inferencing allows AI models to be applied to real-world situations and provide valuable insights, predictions, or automated decisions. Just like the student applying their learned math skills to solve real-world problems, AI models use their learned knowledge to make inferences on new data.

Decentralized Inferencing

Decentralized AI inferencing is a novel approach that distributes inference requests across multiple nodes in a network. The main problem decentralized inferencing solves for is the inferencing privacy and verifiability.

@hosseeb outlines a good example to illustrate why this is important:

Imagine you plan (and have paid for the privilege) to use Llama2-70B to analyze a problem.

When you receive the output, how do you know the output was from the model you paid to use? 

You could’ve received an output from a simpler model, such as Llama2-13B, giving you worse analysis, and pocketing the difference. You would likely be none the wiser.

When it comes to centralized models, you basically have to trust that centralized providers like OpenAI will stay true to their word and provide an output from the model you paid to use. However, in decentralized AI, honesty is not assumed, it is verified.

This is where verifiable inference comes into play. In addition to providing a response to a query, you also prove it ran correctly on the model you asked for. This is achieved through:

• Zero-Knowledge Proofs (ZK ML)

• Optimistic Fraud Proofs (Optimistic ML)

• Cryptoeconomics (Cryptoeconomic ML)

All three techniques - Zero-Knowledge Proofs (ZK ML), Optimistic Fraud Proofs (Optimistic ML), and Cryptoeconomics (Cryptoeconomic ML) - aim to solve the trust and verification problem in decentralized AI inferencing. They ensure that the inference process is executed correctly and that the results are accurate, without relying on a single centralized authority. While each approach has its own tradeoffs in terms of security, cost, and latency, they all share the common goal of enabling trustless and verifiable inferencing in a decentralized network.

Related Projects

Here are some of the projects building in this niche:

1. Allora Network is creating a decentralized marketplace connecting consumers, workers, and validators to produce reliable inferences. The network uses private machine learning models, exclusive subgroups, and on-chain inference commitment to ensure verifiable and high-quality inferences. Their unique reward distribution mechanism fosters a trustless environment for accurate inference generation.

   • Backed by Delphi Digital, Polychain Capital, Framework Ventures and more.

2. Nesa offers a lightweight Layer-1 platform for secure, private AI inference using advanced methods like ZKML and split learning (SL). It provides a decentralized alternative to centralized platforms, featuring a trustless query marketplace powered by a reward economy for AI developers, queriers, miners, and model reviewers.

   • Crowdfunded, no information available on total raise. Partnered with Harvard and has team members with backgrounds in major centralized AI companies.

3. Ritual Net is building a sovereign execution layer for AI, starting with Infernet, enabling developers to access models on-chain and off-chain. By standardizing inference workflows and ensuring AI computations are transparent and reliable, Ritual Net aims to simplify running AI models in a decentralized environment, supporting new applications at the intersection of crypto and AI.

   • Backed by Archetype, DAO5 and many other major VCs and angel investors.

Meme of the Day