1. Who is involved? The story focuses on Eric Zelikman, an AI researcher who left xAI and is currently a PhD student at Stanford University
2. What is the initiative? He is launching a startup called Humans& and aiming to raise US$1 billion in funding, valuing the company at approximately US$4 billion
3. What is the goal? Humans& intends to develop AI models with emotional intelligence (“EQ”) in other words, systems that do more than process language and attempt to understand humans’ emotions, values and long-term goals
4. Why now / significance? The move comes amid a surge of early-stage AI investment, where funders are betting on smaller, talent-rich teams rather than only incumbents. Zelikman argues current models are “too cold and machine-like” and miss longer-term human intent.
In the current wave of artificial-intelligence investment, one researcher is choosing a distinct path. After his departure from xAI in September, Eric Zelikman is pursuing not just another language model, but one built to engage with human emotions. His new venture, Humans&, is reportedly in the process of raising approximately $1 billion at a valuation around $4 billion. The ambition is to craft AI systems that understand not only what people say, but what they mean, what drives them, and how they feel Zelikman’s background anchors this endeavour. At Stanford he led a paper describing how language models can teach themselves to “think before speaking.” That research preceded his technical roles at Microsoft and at Lazard, and then at xAI. Now he proposes to flip conventional AI assumptions: instead of treating each conversational turn as an isolated game, the model should reflect cumulative context, values and objectives. He argues current language models fail to apprehend the long-term implications of statements or actions. He sees an opportunity for AI to become collaborator rather than mere responder Humans&’s angle is two-fold: first, to build models trained to recognize human emotions and goals, and second, to deploy them in contexts beyond chat – the medical sphere (such as cancer research), complex problem-solving, and team collaboration involving mixed human-AI groups. According to Zelikman, achieving major breakthroughs (for example in disease) is more likely if the AI “understands different people’s goals, different people’s ambitions, different people’s values.” He contends that emotional awareness is a missing ingredient in the current state of large language models.
The planned billion-dollar round places Humans& amid the growing list of well-funded research-driven startups in AI. Earlier this year, for example, Thinking Machines Labs raised $2 billion at a $12 billion valuation. The pattern reflects investor belief that breakthroughs will emerge from smaller, focused teams led by top researchers. Zelikman’s timing appears strategic the field is ripe for the next move beyond scale for scale’s sake. If scale alone delivered human-level collaboration or empathy, it might already have been achieved Still, risks abound. Raising $1 billion presumes investors share the vision and believe an emotionally-aware model can deliver commercially viable value and/or societal impact. Skeptics could point to the difficulty of training models to genuinely “understand” humans in a deep sense, and of converting that into robust products or revenue. The valuation assumes not just technical success but significant market adoption For entrepreneurs, founders and investors watching the AI space, Zelikman’s bet offers a noteworthy lesson. The shift from building ever-bigger language models to thinking about how they integrate with human values signals a maturation of the field. Rather than focus purely on parameter counts or compute budgets, the emphasis is shifting toward alignment, human-centred design, and real-world collaboration If successful, the result could reshape how we think of AI assistants, not just as tools but as partners capable of understanding nuance. For you (as an entrepreneur) the implication is clear: opportunities now lie not just in scale, but in depth embedding emotional intelligence, ethics, and value alignment into systems that serve people, teams and goals. The startup also shows that in a crowded market you can differentiate by orientation (empathy + collaboration) as much as by raw technical horsepower.
Humans& is a heavyweight proposal: $1 billion in funding, a $4 billion valuation, top-tier research leadership, and a shift in narrative from “bigger models” to “better-aligned models”. Whether it delivers remains to be seen. But the ambition alone is a signal: the next frontier in AI may not just be machines that talk it may be machines that understand.