Interview, Podcast
a16z Podcast | The IQ and EQ of Robots
- Anki's strategy for developing advanced robotics involves a "bottoms-up" approach, starting with the consumer toy market to build foundational capabilities in manipulation, perception, and navigation before targeting the "holy grail" of home service robots.
- The market for consumer mobile manipulators stagnated after Sony discontinued the AIBO robot dog in 2006, having sold fewer than 200,000 units at a price point of approximately $2,000, due to prohibitive costs and technology limitations.
- Current robotics innovation is driven by leveraging the smartphone supply chain, which has reduced the cost of essential components like cameras, processors, and motors from hundreds of dollars to under $1, making mobile manipulation commercially viable for the first time.
- The "Roomba model" for success in robotics involves identifying a single task a robot can perform sufficiently well to satisfy users, whereas tasks requiring high-fidelity manipulation (e.g., cooking) remain years away from mass-market viability due to reliability and cost constraints.
- Anki's "Cosmo" and "Vector" robots are designed as companion entities rather than purely utility tools, aiming to replicate the emotional dynamics of pet ownership by offering entertainment, personality, and the ability to be physically interacted with without biological maintenance.
- Vector represents a technical leap over Cosmo by being "de-tethered," featuring an onboard quad-core CPU, tablet-like processing power, and cloud connectivity that allows the robot to be always-on, initiate interactions, and function independently of a mobile device.
- The company has released Software Development Kits (SDKs) for its products, including a Python interface for Vector and a graphical interface for Cosmo, to allow users ranging from children to PhD students to program robot behaviors and learn robotics fundamentals.
- Anki employs an internal animation studio with talent from Pixar and DreamWorks to physically animate the robots using Maya software, creating a "personality engine" that outputs reactions based on real-world context and sensor data.
- Optimization of eye contact behavior demonstrated a significant impact on engagement, where increasing the frequency and duration of eye contact by roughly one second raised the average session length from 29 to 40 minutes.
- User testing revealed that children exhibit high empathy toward the robots, with some users intentionally throwing games to avoid making the robot appear "sad" or "upset," a dynamic not typically observed in standard video game interactions.
- Design philosophy prioritizes non-humanoid, "cute" form factors to increase consumer forgiveness for technical limitations and to avoid the "uncanny valley," while using expressive screens and tonal voices to convey emotion without requiring advanced speech synthesis.
- Vector's hardware improvements were directly driven by empathy insights, specifically adding capacitive touch sensors for tactile interaction and a microphone to enable voice responsiveness, addressing the illusion-breaking nature of the previously tethered Cosmo.
- Integration with Alexa is scheduled for December, representing a shift from command-and-control voice interaction to a character-initiated relationship where the robot can proactively seek attention and deliver personalized information.
- Early usage data for the Vector character shows engagement rates of 10 to 12 queries per day, significantly surpassing the industry standard of 1 to 2 queries per day for traditional voice assistants, even before full Alexa functionality is integrated.
- Dave Turetsky notes that interacting with physical robots helps the general public develop "computer literacy" regarding perception and planning, creating a user base capable of understanding complex robotic concepts that were previously abstract.
- Future development roadmaps focus on reducing the cost of depth cameras and specialized AI hardware to below $10, aiming to make 3D environmental mapping and object recognition standard features in consumer products within the next three to five years.
- Anki views the ultimate goal as a catalyst where environmental understanding and interaction capabilities become sufficiently advanced to transform physical domestic tasks into software-driven problems, enabling scalable home utility.