HUTE is an AI research company working on the part of intelligence machines still get wrong: people.

HUTE is an artificial intelligence research and technology company. Its mission is stated in one sentence, and everything the company does is measured against it: making social intelligence and human behavior foundational to artificial intelligence.

The name stands for Human Technology. If you hover over the wordmark at the top of any page, it spells itself out.

What HUTE is, in plain terms

• A research organisation. HUTE studies how people think, feel, communicate, relate and cooperate, and how AI systems can reason about all of that accurately and with appropriate doubt. See the Research Lab.
• An infrastructure company for AI developers. HUTE designs expert-verified, rights-cleared data, environments and evaluations that help others improve how their models handle people. See Intelligence Infrastructure.
• The maker of HUMAN. HUMAN is HUTE's first consumer product: an AI-powered network for human connection, currently in development. See HUMAN.

HUTE is the company; HUMAN is its product. The relationship is like the one between a research company and the product it ships. HUMAN is not the whole of HUTE, and HUTE is not a social app.

The problem

Modern AI systems write fluently, but fluency is not understanding. A model can produce a warm, grammatical reply and still miss that the sender was being sarcastic, that the person it is addressing does not know what the sender knows, or that a polite refusal was really an invitation to try again later. Research on theory of mind in language models shows both striking successes and brittle failures, and there is no consensus yet on what the successes mean; our review of the field sets out the evidence.

Much of what people know about each other is also tacit: it was never written down, so it never reached the data that models learn from. Whole regions, dialects and ways of speaking are thinly represented. And long relationships, in which meaning builds over months and years, are almost invisible to systems that see one conversation at a time. HUTE exists to work on that gap.

Three verbs

• Understanding humanity. Research and evaluation so that AI can reason more faithfully about what people want, how they interpret situations, and how culture and relationships shape meaning.
• Connecting humanity. An AI-powered network that helps real people find, meet and stay in touch with one another. The AI helps with introductions and follow-through; connection itself stays human to human.
• Mobilising humanity. More ways for people to work, share expertise and earn in an AI economy, with people in control of how they are found, represented and paid.

Principles

1. Human agency. People control how they are found, contacted, represented and compensated.
2. Consent is specific. Using a helpful product is not consent to become training data. Private HUMAN conversations, contacts and memories are not research or commercial training data without separate, explicit authorisation and the rights to give it. Read more on our Responsible AI page.
3. Evidence over claims. We label every research output by evidence status, we report uncertainty and disagreement, and we separate what is built from what is planned. See our methodology.
4. No manufactured credibility. We do not claim customers, partnerships, benchmarks or products that we cannot show.

HUTE at a glance {#facts}

Read next

• What is social intelligence in artificial intelligence?
• Research Report 001: the state of social intelligence in AI
• Meet the founders