Geo Founder: Human Verification Can Fix AI’s Internet Data Problem

March 5, 2025 — Geo founder Yaniv Tal told crypto.news that unreliable AI answers stem from flawed internet data, not model design, arguing that lost provenance, flattened authority, hidden disagreement, and repeated model-generated errors undermine AI reliability. Tal’s proposed solution separates claims from sources within community-governed Spaces, using human judgment to rank credible reasoning while preserving competing viewpoints.

Immediate Details & Direct Quotes

Tal identified four specific weaknesses in online information that compromise AI answer quality. “AI doesn’t have a truth problem, the internet does,” Tal said. The first issue, lost provenance, occurs when claims get scraped and restated across multiple pages before entering training data, making original sources unrecoverable. “Provenance collapses. A claim gets scraped, restated, and re-scraped until the original source is unrecoverable,” he explained.

The second weakness involves flattened authority, where research papers, company announcements, and anonymous forum posts receive equal weight in training pipelines. Hidden disagreement represents the third problem—language models compress competing expert positions into single confident answers without acknowledging credible alternatives exist.

The fourth issue involves model-generated content re-entering training data. “Models increasingly train on output from other models, so errors don’t just persist, they amplify,” Tal said. A 2024 Nature study confirmed this risk, documenting “model collapse” where systems gradually lose information about original data distributions.

Market Context & Reaction

Geo’s proposed solution centers on separating statements from authors and supporting evidence, creating structured relationships showing which arguments support, contradict, or respond to disputed points. This structure ranks arguments by assessed strength while retaining competing material beneath them. “Pluralism doesn’t have to mean noise,” Tal said. “We think several well-structured competing perspectives is often the more honest answer.”

Geo organizes information through independent communities called Spaces, covering crypto, health, AI, education, world affairs, and US politics. The model derives from Geo Genesis, which entered early access in January 2025 with Editors holding voting authority and Members contributing information. Built on Aragon OSx governance framework, Geo Genesis followed GRC-20, a standard for representing connected knowledge on-chain.

Tal, who co-founded The Graph before starting Geo, said expertise would not come from central authority. Contributors build reputation through their work, with records following individuals between Spaces. “Human judgment is the scarce input now, not the redundant one,” Tal said.

Background & Historical Context

The National Institute of Standards and Technology’s Generative AI Profile, published July 2024, recommends controls overlapping with Tal’s argument—documenting training-data sources, monitoring generated material origin, and incorporating domain expert feedback. The Coalition for Content Provenance and Authenticity uses cryptographically signed Content Credentials to preserve digital asset origin information.

Blockchain serves as Geo’s first Space because it tests verifiable records against human interpretation. “On-chain data is verifiable by construction. A transaction happened or it didn’t,” Tal said. “Everything wrapped around it is claims.” Project announcements, partnership descriptions, and market forecasts require separate treatment attached to named contributors with preserved records, including earlier incorrect statements.

Community governance faces identity challenges including Sybil attacks, where participants create multiple identities to gain influence. Tal said Geo relies on contribution records and domain-specific communities rather than allowing anonymous material equal status with claims from people holding public histories.

What This Means

Tal’s approach suggests shift toward structured knowledge verification in crypto markets where promotional activity often mixes with verifiable transaction data. Geo intends to label categories separately rather than presenting all information in the same voice.

The framework implies ongoing development of community-governed information systems that preserve disagreement while ranking argument strength. “Skin in the game without financial stakes changes the quality of what people are willing to put their name to,” Tal said.

Contributors seeking involvement can apply as editors through individual Spaces, with members participating in governance of subjects they follow. The long-term question remains how independent Spaces prevent coordinated manipulation and whether expertise transfers across subjects.

This approach aligns with broader crypto security trends where manual verification complements automated tools. As AI-generated content proliferates, systems that track provenance and maintain human accountability may become increasingly valuable for traders seeking reliable information.

—