No Priors• Oct 9, 2025• 36mInterview

No Priors Ep. 135 | With Humans& Founder Eric Zelikman

From No Priors

Eric Zelikman•Founder, Humans&

Executive Summary

Eric Zeltman, formerly of Stanford and XAI, has founded a new company, Human's End, to shift the focus of AI development from pure 'IQ' to 'EQ' (emotional and interactive intelligence).
He argues that the prevailing industry paradigm is too task-centric and focused on automating humans out of the loop, which limits innovation and human potential.
Human's End aims to build models that deeply understand long-term user goals, context, and values, viewing this as the most fundamental missing capability in AI today.
Zeltman's philosophy is that AI designed to empower and collaborate with people can grow the overall economic pie, whereas AI designed simply to replace human tasks will only redistribute it.

12 quotes

Concerns Raised

The AI industry is overly focused on task-centric benchmarks and automating humans out of the loop.
Current models fundamentally lack 'EQ,' including long-term memory and a deep understanding of user goals.
An AI strategy focused on replacement will merely carve up the existing economic pie, limiting true innovation.
The trend of models working autonomously for longer periods could lead to outputs that humans don't understand or trust.

Opportunities Identified

Developing AI with high 'EQ' to understand and collaborate with users is a major untapped area for innovation.
Building models that empower people can grow the overall economic pie and unlock latent human potential.
There are still meaningful, unexplored dimensions for scaling AI capabilities, especially by incorporating humans into the development loop.
A serious effort in building AI with memory and user-modeling capabilities could yield significant performance improvements over current systems.

Key Themes

The Limits of 'IQ'-Centric AI

The discussion critiques the industry's heavy focus on scaling raw intelligence ('IQ') through task-centric benchmarks. Zeltman acknowledges that models are becoming smarter on verifiable tasks but argues this approach neglects more fundamental capabilities and creates a 'jagged' performance profile.

This highlights a potential blind spot in the current AI race, suggesting that focusing solely on performance metrics like MMLU or coding benchmarks may lead to powerful but ultimately brittle and unhelpful systems.

Human-in-the-Loop vs. Full Automation

A central theme is the philosophical and practical choice between designing AI to replace humans versus empowering them. Zeltman asserts that the default instinct of most labs is to remove the 'messy' human element for easier scaling, but that keeping people in the loop is a crucial, active design choice for achieving superior outcomes and innovation.

This debate shapes the future of work and the economic impact of AI. Companies must decide whether their AI strategy is one of substitution or augmentation, which has profound implications for product design and market creation.

The 'EQ' Frontier: AI's Next Capability Axis

Zeltman posits that the next major frontier for AI is not just reasoning but 'EQ'—the ability to understand user goals, emotions, values, and long-term context. He frames this not as a feature for companionship apps, but as a core productivity-enabling capability that is essential for building truly helpful AI.

This signals a shift from task-completion agents to collaborative partners. Developing AI with 'EQ' could unlock new product categories and dramatically improve user adoption by making AI more intuitive, adaptive, and aligned with human intent.

Beyond Task-Centric Training

The current AI training paradigm, focused on discrete tasks, is identified as a key reason models lack crucial features like long-term memory and a deep understanding of user intent. Zeltman advocates for a new approach where the model's objective is to build a persistent, evolving model of the user.

This technical critique points to a necessary evolution in AI R&D. Progress in areas like personalization, agentic workflows, and proactive assistance depends on moving beyond the limitations of the current training methodology.

The Evolution of AI Reasoning

The conversation touches on Zeltman's foundational research, including the Star and QuietStar algorithms. These methods, which involve models iteratively generating and learning from their own solutions, represent key steps in improving AI reasoning and demonstrate a path to scaling these capabilities.

Understanding the principles behind modern AI reasoning techniques is crucial for appreciating both their power and their current limitations. This context explains how models improved from simple pattern matching to solving complex, multi-step problems.

Get started free

Topics

AI Reasoning Reinforcement Learning (RL)Star Algorithm QuietStar AI Scaling Human-in-the-Loop AI Alignment Emotional Intelligence (EQ)User Modeling AI Agents Task-Centric Benchmarks Long-Term Memory Human-Computer Interaction Future of Work Human's End Eric Zeltman XAI

Processed Mar 31, 2026 yt-dlp + mlx-whisper + Gemini