Robots
Menlo is experimenting with using today’s AI advances to build generally intelligent robots, with world understanding (e.g. physical, spatial) and reasoning capabilities.
- Most robots today remain expensive piles of limbs
- Execute predefined tasks in controlled environments
- Fail when faced with the complexity of the real world
We are focused on the “higher order brain” of the robot, and work with popular robot brands for the robots and their SDKs.
Goal
We are working on the following:
- Sensors + Hardware to enable robots to understand the world
- Reliable control, management and telemetry systems for robots (for both Teleop and AI)
- Energy-efficient hardware to power on-device AI
- Latency-minimized intelligence using combination of on-device and cloud AI
This workstream also guides Menlo’s existing work on:
- Multimodal models with world understanding
- Continuous learning
- Memory
- Problem solving and reasoning
- Speech and human-robot interaction
Long-Term
Robots are a long-term endeavor, and Menlo’s current view is that there we are missing big pieces. Fundamental breakthroughs needed over the next 5-10 years at both the hardware and software levels. Even then, cost needs to drop by a factor of 10x to make mass robots possible.
However, the next great leap is intelligence that can act in the physical world. Embodied systems that understand and manipulate reality, will fundamentally transform the potential of human civilization.
Menlo continues to invest in this, as we strongly believe in its transformative potential:
- Wheels transformed human’s mobility
- Writing transformed human’s civilizational memory
- Computers transformed human’s communication and info processing
We strongly believe robots will transform human civilization’s ability to master our physical world, and allow us to move towards a Kardashev Type-1 civilization .