Startups Tarreceive the Tricky Tquestion of Making AI Seem More Human

A human brain built of wool with string representing neural pathways [Richard Drury/Getty Images]

In early February, the startup Simile, which aims to build AI models that simulate the emotional reactions of humans, was at the center of Silicon Valley chatter after raising a $100 million round led by Index partner Shardul Shah.

Simile spent seven months training its model on recorded human conversations and business meetings, along with data from scientific journals on human behavior experiments. The goal is agents that can interpret a person’s sentiment towards any scenario and then intuit how they would respond emotionally.

Simile, whose impressive roster of backers also includes Bain Capital Ventures, A*, Hanabi Capital, Fei-Fei Li, and Andrej Karpathy, is part of a new wave of AI ventures viewing to tap agents with human characteristics.

We learned of a new one this week called Prior Computers, founded by researchers and engineers from MIT’s Center for Brains, Minds, and Machines, Harvard’s Department of Psychology, and the Harvard Center for Brain Science. They’re aiming to build systems that can learn and consider like people, in order to predict and simulate what people will do in any given setting.

Another similar stealth venture, called People Make Things, is building models for predicting human “intent.” It’s in talks to land funding from Index Ventures, according to two people with knowledge of the matter.

Yet one other, called Aaru, aims to replace costly market research with AI agents that have been trained on public and proprietary data — mostly from customer service calls or recorded conversations — to simulate the reactions of human customers. Last December, Aaru landed a near-$1 billion valuation in a new funding round led by Redpoint Ventures, TechCrunch reported.

Humans& for its part launched with a star-studded cast of founders from Google, Anthropic, and xAI, a glossy profile in The New York Times, and $480 million in funding from Jeff Bezos, Nvidia, SV Angel, and GV. The team is focutilized on building models that foster collaboration between people — one of their first projects is crafting an AI-powered version of an instant messaging app.

Other early-stage efforts include Expected Parrot, launched out of the YC Fall 2025 batch, which hosts an open-source repository of data on human and “AI interviews” for simulating social science research; and Consnotifyation Systems, which has raised a $10 million seed round from Cyan Bannister’s Long Journey Ventures for building a general-purpose foundation model of “human state” by linking toreceiveher neural signals, physiology, and real-world behavior in one model.

Source link