Conference Presentation, Keynote, Product Demonstration
Nils Pihl, CEO & Co-Founder, AUKI Labs: Teaching Machines to See the World
- Projected Economic Expansion: The speaker forecasts a minimum 10x growth in the global economy over the next 20 years, driven by a rise to approximately 100 billion productive economic participants, the majority of which will be Artificial Intelligence.
- This projection relies on the assumption that current technology will accelerate productivity growth by roughly three times the rate observed over the last two decades.
- Historical context notes a 3x growth in the last 20 years and a 10x growth since the mid-1980s.
- The Physical Economy Gap: While software has largely transformed the white-collar sector, approximately 70% of the global economy remains tied to physical locations and labor where current AI lacks direct utility.
- Current LLMs cannot perceive real-time physical environments (e.g., chair arrangement, retail space dynamics) due to a lack of physical presence.
- Future economic value unlocking requires AI to access and interact with physical spaces, not just digital ones.
- Coordination Protocol Deficit: A 2014 analysis by Naval Ravikant identified a missing "fifth protocol" in the internet stack required for autonomous agents (robots, self-driving cars) to coordinate by exchanging economic value.
- Without this protocol, autonomous systems cannot negotiate physical interactions, such as resolving lane merges or traffic coordination.
- Limitations of GPS and Location Services: Global Positioning Systems (GPS) are increasingly insufficient for high-density urban robotics due to their reliance on line-of-sight technology.
- GPS accuracy degrades in major metropolises with high-rise density, such as New York (900+ buildings over 100m) and Hong Kong (4,000+ buildings over 100m).
- The speaker cites the Greater Bay Area (87 million people, smaller than Greater Los Angeles) as a primary example of where GPS failure is imminent due to building density exceeding that of all of North America and Europe combined.
- Traffic inefficiencies in cities like Beijing result in a loss of human productivity roughly equivalent to "one pyramid's worth of time" every week.
- Six-Layer Robotics Stack Requirements: Mass deployment of humanoid robots requires a software stack comprising six distinct layers:
- Locomotion: Basic movement and balance (currently achieved by companies like Unitree).
- Manipulation: Physical interaction with objects (e.g., pouring, grabbing); this layer remains underdeveloped.
- Spatial Semantic Perception: Distinguishing object identities (e.g., differentiating a plant from a cat).
- Mapping: A dynamic context layer independent of perception weights to handle a changing physical world.
- Positioning: Localization relative to the map, replacing or augmenting GPS.
- Applications: The interface layer tying all functions together.
- Aukey's Strategic Approach: Aukey focuses exclusively on layers 3 through 6, avoiding locomotion and manipulation by leveraging human labor and existing hardware (smartphones, smart glasses) for movement.
- The company treats the iPhone as a robot "head" to accelerate AI perception, mapping, and positioning development.
- A recent pilot with a European retailer used smartphone camera data to create digital twins, enabling product placement analytics.
- Quantifiable Retail Outcomes:
- Staff walking distance reduced by 40% via AR search and optimization.
- Employee time savings of at least 15 minutes per day through AI and AR task management.
- Hardware Form Factor Strategy: The speaker introduces programmable smart glasses (developed with partner Mentra) weighing 40 grams, based on historical data that wearable devices over 40g tend to be abandoned.
- This hardware facilitates AI perception, mapping, and positioning while relying on human locomotion and manipulation, described as "hybrid robotics."
- Decentralized Physical Infrastructure Networks (DePIN): The speaker proposes a decentralized network of self-hosted spatial maps as the successor to GPS, termed a "universal spatial computing protocol."
- This architecture allows devices (phones, robots, glasses) to share and compensate for spatial data without relying on a central cloud provider.
- Privacy concerns drive this model: retailers and individuals require the ability to self-host visual merchandising and location data rather than trusting centralized clouds.
- Privacy and Data Sovereignty Concerns: Current trends in AI hardware creation pose significant privacy risks.
- Elon Musk noted on the Lex Friedman podcast that the Optimus robot could become "the greatest source of data in the world."
- Smart glasses manufacturers are increasingly distributing hardware at a loss to capture continuous visual data from users.
- Enterprise clients (e.g., retailers) often cannot store visual merchandising data in the cloud due to proprietary sensitivity.
- Future Outlook: The speaker envisions a shift where AI accesses a physical world made accessible through decentralized infrastructure.
- Autonomous robots will navigate by connecting to local municipal or corporate maps (e.g., a robot entering Carrefour connects to Carrefour's map) rather than a single global map.
- The market is posited as "the first artificial intelligence humanity ever invented," capable of efficiently allocating hardware resources based on demand.
- The ultimate goal is to treat the physical world as the new frontier for AI expansion, unlocking economic potential where labor is no longer a constraint.