AI headlines:- OpenAI launches ChatGPT Atlas, an AI browser targeting Google- ChatGPT Go offered free in India- Reddit sues Perplexity over content scraping- Tencent reports 50% of new code AI-generated- AWS outage highlights cloud vulnerabilitiesMeanwhile, we boosted Expander bandwidth from 1 to 6 TB/s verifiable.Proof at scale: turning verifiable computation into a high-speed reality.As AI accelerates, verifiability ensures alignment.That’s why we’re building zkML, zkGPT, and Polyhedra ID.Speed with proof ...

AI at scale isn’t just fast—it’s sloppy.– $440K Deloitte AI report full of errors– Claude rolling out to 470k staff– Stanford: “work slop”– Altman: oceans of slop are comingThe solution isn’t more AI.It’s verifiable AI.zkML. zkGPT. Polyhedra i-D.#BlackBox https://t.co/9i9fhN9UJ4 ...

AI just leapt out of the screen:– OpenAI building its own device– Meta’s glasses read your muscle signals– Harvard reversing disease cells– Albania appoints an AI as a ministerBut if AI runs our bodies & governments—how do we trust it?That’s why we’re building zkML, zkGPT, and Polyhedra ID. ...

AI Development & Future - The future of AI is verifiable, indicating a focus on transparency and trust in AI systems [1] - zkGPT prototype and repository are coming soon, suggesting rapid development and open-source initiatives in the AI field [1] - Integration with zkPyTorch is planned, highlighting the importance of compatibility and collaboration within the AI ecosystem [1] - zkML-as-a-service is on the horizon, pointing towards the commercialization and accessibility of AI technologies [1]

2/Why zkGPT matters:- Stops providers from downgrading models secretly- Enables regulator-ready AI- Protects model IP with proof-as-watermark- Unlocks trust for enterprises & high-stakes use cases ...

Join @sampearton and @jiahengzhang96 on The Polyhedra Show for an introduction to zkGPT on Monday 18th Aug at 7am UTC.Verified community members engaging will win USDT. Make sure to join our Discord to be eligible for rewardshttps://t.co/uGLwFzaD5p https://t.co/17JBnxasPp ...

1/🧵diving deeper into our new work on zkGPT: Proving LLM inference fast with Zero-Knowledge Proofs.Why? Service providers might deploy a smaller/cheaper model than promised. ZK lets them prove correctness without revealing model parameters.📄 https://t.co/ToX51Wla56 ...