Frequently asked questions

Direct answers to what people most often ask about AI USM: what the technology does, what it deliberately does not do, and how data is handled.

What is Emotion AI?

Emotion AI is artificial intelligence that estimates a person's emotional context from observable signals — facial dynamics, voice prosody and language — and uses that estimate to shape how it responds. It is the applied side of the research field known as affective computing.

What is emotion recognition AI?

Emotion recognition is the measurement step inside Emotion AI: turning a camera, microphone or text stream into an estimate of emotional state with a confidence attached. AI USM performs it across three channels at once rather than relying on any single one.

How does Emotion AI work?

Each input channel is analysed separately — expression dynamics from vision, pitch, tempo and pauses from voice, intent and hedging from text. A fusion layer weighs the channel estimates against one another and produces one emotional context, which then informs the wording, the avatar's expression and what is stored in memory.

What is multimodal emotion recognition?

Combining several signal channels into one estimate. Single channels fail in different, predictable ways — bad lighting, background noise, terse typing — so reading vision, voice and text together makes the result more robust than any one of them alone.

What is affective computing?

Affective computing is the research field concerned with systems that recognise, interpret and simulate human affect. Emotion AI is the commercial name for the applied technology; emotion recognition is one task within it. AI USM works on the applied end of the field.

What is empathetic AI?

Empathetic AI is an interactive system that takes emotional context into account when it responds — adjusting tone, pacing and framing — rather than answering only the literal request. It does not mean the system feels anything.

How is Emotion AI different from sentiment analysis?

Sentiment analysis labels text on one axis: positive, negative or neutral. Emotion AI works across modalities and a richer state space, and it can detect a mismatch between them — a calm sentence spoken in a strained voice is exactly the case sentiment analysis cannot see.

Which modalities can AI USM analyse?

Live camera input, speech audio, typed and spoken text in five languages, plus uploaded images and documents used as context. Consent is given per modality: voice analysis can be allowed while the camera stays off.

How can Emotion AI be used in healthcare?

As decision support: emotion-aware intake conversations that record how a symptom was reported alongside what was reported, follow-up that notices disengagement, and summaries that a clinician reviews. AI USM does not diagnose and is not a medical device.

What are the limitations of Emotion AI?

It estimates emotional context from observable signals; it does not read minds and it can be wrong when a channel is degraded. It should not be used to score, screen or rank people, and an estimate should never be treated as a fact about someone.

Is Emotion AI the same as emotional intelligence?

No. Emotional intelligence is a human capacity. Emotion AI is software that measures signals correlated with emotion and adapts its behaviour accordingly. The system has no feelings and makes no claim to have them.

Are these case studies customer references?

No. They describe how the platform is designed and deployed in pilots and in-product features. We do not publish customer names, metrics or testimonials that we cannot evidence.

Can AI USM be used to diagnose a medical condition?

No. AI USM is not a medical device and does not provide a diagnosis. Health-related assistants provide information and structure a conversation for a qualified clinician to review.

How accurate is the emotion recognition?

The fused multimodal model reaches high accuracy on our internal validation sets — higher than any single channel on its own. Accuracy varies by channel quality: poor lighting or a noisy microphone lowers the contribution of that channel, and the fusion layer weights it down accordingly.

Which languages are supported?

The interface and assistants work in English, Russian, Chinese, Czech and Arabic. Voice and vision analysis are largely language-independent; text understanding is trained per language.

What happens to camera and microphone data?

Raw video and audio are used for inference and are not retained by default. What persists is the derived, anonymised emotional state, and only where the user has consented to that modality.

Where is the company registered?

Two entities: USM Tech Inc., a Delaware (USA) corporation, and AI USM LTD, registered in ADGM Abu Dhabi under number 28022 and endorsed by Hub71.

What is the ULC Token used for?

ULC is a Solana SPL token used for assistant access, partner billing and ecosystem incentives. Details are on the /token page and in the whitepaper.