New frontiers in Human AI Interface

Introduction

For two decades, our interaction with the digital world has been confined to a mere 5-inch screen and a single fingertip. But what if we could break free from these limitations and unlock the full spectrum of our innate human senses in our everyday computing?

The past few years have witnessed a dramatic acceleration in the development of human-AI interface technologies, pushing the boundaries of how we interact with artificial intelligence. From immersive display technologies to intuitive wearable devices and ambitious AI-powered assistants, the landscape is rich with both groundbreaking innovations and valuable lessons from early attempts.

Recent Announcements and Key Players:

Meta Connect Announcements: Display Glass and Neuron Wristband

Meta’s annual Connect event has consistently served as a platform for showcasing their long-term vision for augmented and virtual reality. The introduction of “Display Glass” hints at a future where digital information seamlessly blends with our physical world, likely offering contextual overlays and interactive experiences without the bulk of traditional headsets. Complementing this is the “Neuron Wristband,” suggesting an advanced input method that could potentially interpret neural signals or subtle hand gestures, offering a more natural and less intrusive way to control devices and interact with AI. These developments underscore Meta’s commitment to building the foundational hardware for the metaverse, where human-AI interaction will be paramount.

Apple’s AirPods Pro with Live Translation

Apple’s iterative approach to innovation often involves integrating advanced AI capabilities into their widely adopted ecosystem. The “Live Translation” feature in AirPods Pro is a prime example of this, leveraging on-device and cloud AI to break down language barriers in real-time. This not only enhances communication but also demonstrates the potential for AI to act as a personal, omnipresent interpreter, seamlessly facilitating interactions in a globalized world. It highlights a focus on practical, everyday applications of AI that enhance user experience without requiring entirely new form factors.

Google’s Continued Effort in Smart Glasses

Google has a long history with smart glasses, from the ambitious but ultimately limited Google Glass to more recent enterprise-focused solutions. The “continued effort” suggests a persistent belief in the potential of head-mounted displays as a human-AI interface. Future iterations are likely to focus on improved form factors, enhanced AI capabilities for contextual information delivery, and more robust integration with Google’s vast array of services, including search, maps, and AI assistants. The challenge remains in finding the right balance between utility, social acceptance, and privacy.

OpenAI Acquires IO (Jony Ive)

OpenAI’s acquisition of “IO,” a design collective led by former Apple chief design officer Jony Ive, is a significant strategic move. This signals a strong recognition within the leading AI research organization that the physical embodiment and user experience of AI systems are crucial for their widespread adoption and impact. Ive’s legendary focus on minimalist design, intuitive interfaces, and emotional connection to technology suggests that OpenAI is not just focused on developing powerful AI models, but also on crafting elegant and human-centric ways for people to interact with them, potentially leading to new categories of AI-powered devices and interfaces.

Learning from Early Efforts:

Failed experiments? Human AI Pin and Rabbit R1

The Humane AI Pin largely failed due to a combination of technical shortcomings, a high price point, and a flawed value proposition. The device was criticized for being slow, unreliable, and prone to overheating. Its primary interface, a laser-projected screen on the user’s palm, was found to be finicky and difficult to use in bright light.

Furthermore, the $699 price and a mandatory $24/month subscription fee were deemed exorbitant for a device that could not reliably perform basic tasks and lacked integration with common smartphone apps and services. Ultimately, the AI Pin failed to solve a significant problem for consumers and was widely seen as an inferior, redundant gadget compared to the smartphones already in their pockets.

The Rabbit R1’s failure can be attributed to its inability to deliver on its core promises and its fundamental lack of purpose. The device was heavily marketed as a “Large Action Model”-powered tool that could control apps and services on the user’s behalf, but at launch, it only supported a handful of apps and failed at many basic tasks. Reviewers noted poor battery life, sluggish performance, and an awkward user interface.

The company’s claim that its device was not just a smartphone app was undermined when it was revealed that the entire interface ran on a single Android app, raising the question of why dedicated hardware was even necessary. The R1’s limited functionality, combined with its inability to compete with the capabilities of a modern smartphone, led many to conclude it was little more than a “half-baked” toy that did not justify its existence.

Looking Ahead in this article:

The evolution of human-AI interfaces is a dynamic field characterized by rapid experimentation and continuous refinement. How do you keep up with the latest developments and stay a step ahead of the curve?

In the following chapters, we will start with a deep dive into the human-machine interface in the context of AI. This will be followed by an opportunity analysis for future AI-focused HMI, as well as an overview of 40 companies categorized by the senses they address. Hopefully, this will give you a bird’s-eye view of this fast-progressing industry and provide you with a roadmap to explore further according to your personal interests.

Human Machine interface — a deep dive

Comparative Table of Human Senses for HMI

Key Observations

1. Vision dominates bandwidth — orders of magnitude higher than other senses, but also easiest to overload (cognitive bottleneck at ~40–60 bps for conscious reading/listening).
2. Latency matters differently: vestibular & proprioception are fast reflexive senses — latency below ~20 ms is essential, otherwise motion sickness / disembodiment occurs. Vision tolerates 50–100 ms in UX.
3. Electronic acquisition:

• Easy: vision (cameras), hearing (mics).
• Medium: touch (arrays of pressure sensors, haptic actuators).
• Hard: vestibular (feedback impossible without invasive or rotating rigs), proprioception (requires multimodal sensing), smell/taste (chemical).

4. Importance for HMI:

• Core: Vision, Hearing, Touch, Proprioception, Vestibular.
• Niche / emerging: Smell, Taste, Interoception, Thermoception.
• Critical distinction: input vs output — we can sense vision & hearing easily, but delivering feedback in touch/haptics & vestibular is much harder.

HMI Sensorium Radar

• Vision dominates in bandwidth & importance, with medium acquisition difficulty.
• Hearing offers excellent latency and easy acquisition.
• Touch + Proprioception have high importance but are technically hard to digitize.
• Vestibular scores high on latency sensitivity but is very difficult to reproduce electronically.
• Smell & Taste sit at the low-bandwidth, high-difficulty, low-importance corner (niche).
• Interoception & Thermoception fall in between — valuable mainly for health or immersive feedback.

HMI opportunity map

Implications for AI-HMI Design

• AI Interfaces Today (short-term): vision + hearing dominate (AR glasses, voice agents), but gesture, touch, micro-movements are the new frontier.
• Near-term Breakthroughs: haptics (Afference neural haptics, HaptX gloves), silent-speech (AlterEgo), proprioception mapping (IMU + EMG), vestibular tricks (electro-stimulation).
• Far-term: smell/taste/interoception → highly niche but can create hyper-immersive XR or health-aware AI companions.
• Bottleneck: humans can’t consciously process anywhere near the sensory raw bandwidth — HMI design must compress to what’s useful, intuitive, and low-latency.

HMI Opportunity Map

• Bottom-left (Vision, Hearing) → high bandwidth, low acquisition difficulty → already well-covered, but incremental AI/UX improvements matter.
• Top-right (Vestibular, Proprioception, Touch) → high bandwidth/importance but hard to acquire electronically → biggest innovation opportunities.
• Smell & Taste → low bandwidth, very hard, low importance → niche applications only.
• Interoception & Thermoception → moderate niche, especially for health-aware or immersive HMIs.

The “sweet spot” for future startups lies in making hard-to-digitize senses (touch, balance, body sense) usable for AI interfaces — biggest gap between potential value an`d current tech maturity.

Biggest under-served opportunities for HMI innovation:

I’ve ranked the senses by Innovation Gap (the difference between their theoretical potential and today’s opportunity score).

1. Vision — already dominant but still leaves the largest gap (AI-driven compression, semantics, and augmentation).
2. Proprioception — huge potential but very hard to capture; unlocking it could transform XR/robotics.
3. Touch — high payoff if electronic haptics and tactile sensing improve.
4. Hearing — strong today but still under-optimized (spatial, multimodal, selective hearing AI).
5. Vestibular — critical for immersion, but remains technically difficult.

40 promising HMI startups to watch

Here’s a curated, up-to-date landscape of AI-hardware startups building new human–machine interfaces (HMI) for the AI age. I grouped them by interface modality and flagged the form factor, what’s new, and stage. I focused on 2024–2025 developments and included links/citations so you can dig deeper fast.

First let’s have a look at the AI HMI startup landscape in terms of bandwidth vs acquisition difficulty:

1) Silent-speech, neural/nerve & micro-gesture input (non-invasive)

2) Non-invasive neurotech / everyday BCI wearables

3) Minimally-invasive & invasive BCI (clinical first, consumer later)

4) AR glasses, AI wearables & spatial computers (new UX canvases)

5) Eye-, face- & driver-state sensing (affect-aware, context-aware UX)

6) Quick Watchlist (emerging / adjacent)

• Wispr — software-first but explicitly pitching a voice-native interface for the AI era; raised to build the “keyboard replacement” with AI editing. (Good bellwether for voice-as-primary UX.) (Wispr Flow)
• MindPortal — fNIRS “thought-to-AI” headphones; early but notable. (mindportal.com)
• Ultraleap / Leap Motion legacy — hand-tracking pivot; signals consolidation in the category. (UploadVR)
• HaptX — industrial-grade microfluidic haptic gloves; training robots/AI with rich human demonstrations. (HaptX)

The Road Ahead

This article explores the shift from traditional screen-based HMIs to multi-sensory interactions. It highlights advancements by Meta, Apple, Google, and OpenAI, alongside lessons from past experiments like Human Pin and Rabbit R1. A detailed analysis of human senses considers information transfer, latency, acquisition difficulty, and HMI importance. The “HMI Opportunity Map” identifies innovation gaps in under-digitized yet crucial senses like touch and proprioception. Finally, it lists 40 promising HMI startups categorized by interface modality.

As you navigate these new frontiers, consider how your own work, research, or investments can contribute to creating more intuitive, ethical, and truly human-centric AI interactions. We encourage you to explore the highlighted startups, delve into the cited research, and actively engage in the ongoing dialogue about shaping a future where technology seamlessly augments human capabilities across all senses.