EQTY Lab

We’re big fans of RISC Zero which allows code to be run verifiably. Instead of trusting a cloud provider, a zkVM provides cryptographic proof to prevent tampering during execution. Today, we’re excited to release two tools: 1) An open source web verifier that allows anyone to verify an execution receipt: https://lnkd.in/gwxeXDrz As EQTY Lab works on creating trusted AI solutions, we are seeking new ways to harden AI supply chains. zkVMs have a big role to play. Today we’re hardening AI data preparation using Risc Zero to run verifiable tasks like PII redaction and hash equivalence assertions. In the future, we’ll see more complex verifiable computation pipelines for LLMs and beyond. It’s just the beginning of the story and we want to play our part. Risc Zero receipts include data that the program committed to, but the format is specific to the program being proven so it can’t be generically parsed. So we added a parser interface that any program author can implement to construct a human readable statement. We found this useful for increasing trust that ZK programs were run as expected so we’re open sourcing it. There’s a minimal verifier npm package, a customisable react component npm package, and a hosted version. and 2) an open source web verifier?for running Python code in the zkVM: https://lnkd.in/gqXxCYrJ For Python, you’ll get an execution receipt alongside the Python program outputs. Hop over to our web verifier to check the receipt: https://lnkd.in/gwxeXDrz?

We've spent 18 months developing the AI Integrity Suite to introduce new transparent and accountable tools for open source AI. It’s not a simple future, but one we’re rolling up our sleeves to make happen —responsibly. Accenture's depth in Gen AI and their methodology for security and compliance for public and private sectors make them the perfect place to turn to for testing and evaluating solutions like ours. We’re excited for the path ahead.

The intersection of #DLT and #AI carries significant potential for real-world application. Open-source AI platform ClimateGPT is bringing trust, transparency, and accountability to the world of climate action using the immutability and unmatched energy efficiency of the Hedera network. Recently previewed at #WEF24 in Davos by Jonathan Dotan - Co-Founder of EQTY Lab and Founder of the Starling Lab for Data Integrity - ClimateGPT aims to address the impact of climate change through AI-driven insights and support informed decision-making for those driving towards a greener and more sustainable future. The Banker | https://lnkd.in/ewN_Wu6p

ClimateGPT is the first task-specific ensemble of AI models dedicated to addressing climate change impact. We’re excited to be a partner in training these models via the Endowment for Climate Intelligence (ECI), which is making them publicly available today via Hugging Face. Governed by ECI, ClimateGPT was created by AI experts at Erasmus.AI AppTek and EQTY Lab. This open source AI solution is available in 20+ languages and comprises five AI models, including a groundbreaking 7b parameter foundational model trained on 300b climate-specific tokens. All were computed and hosted entirely with renewable energy. ClimateGPT was trained on over 10b web pages & hundreds of millions of academic articles covering climate technologies and the local impact of extreme weather. It’s trained to synthesize interdisciplinary research & break silos to generate solutions. Find the models and model cards on Hugging Face and the press release on Medium below, which includes a link to the research paper. https://lnkd.in/gegMBrQQ https://lnkd.in/gSBUTd4k

EQTY Lab designs transparent and accountable tools for Responsible AI. We're excited to announce our first tool on Hugging Face: the AI Lineage Explorer — a solution that quickly visualizes the provenance of your models. AI Integrity begins here. EQTY Lab is releasing the AI Lineage Explorer natively on Hugging Face to empower users to see and trust a model’s data sources, architecture, computation, and governance. This level of transparency is crucial for anyone committed to Responsible AI. We create AI integrity with tamper-proof credentials and anchors that attest to each step of the AI training process. Build trust from ingestion to inference with W3C standards and Hedera IPFS Want to see the AI Lineage Explorer in action? Check out the full AI lifecycle of ClimateGPT in the Hugging Face example below — a brand new foundational LLM we trained on green energy to address the impact of climate change. The AI Lineage Explorer is only the first of many tools in EQTY Lab’s AI Integrity Suite that will be released this year. Sign up for early access to our open-source SDK by visiting https://bit.ly/lineagesdk https://lnkd.in/gEct5vKC #AI #HuggingFaceSpaces #Innovation #Transparency #AICommunity