Interview
a16z Podcast | Beyond CES: Connected Home Devices, Voice, and More
- Over the next five years, board-level strategies for large companies will be distinguished from non-essential applications or bureaucratic actions that may fail.
- For pure tech companies, the strategic plan over the next 20 years involves connecting all electrical home appliances, such that items like washing machines become digital or connected.
- Internal organizational dynamics at large firms may lead to three, four, or five different groups attempting to build solutions for a single emerging product trend.
- Future appliances may include screens to perform automated functions, such as detecting load volume to dispense detergent without user input.
- There is an unresolved debate regarding whether non-compute devices like fridges and washing machines require full operating systems and complex interfaces or should remain simple single-purpose devices.
- Companies may adopt a "checkbox minimal API" approach for edge devices to ensure compatibility with competitors, while mandating specific features like Bixby for core products at the board level.
- Product strategies may differ by market, selling unified ecosystems domestically while including API access for competing platforms like Alexa and Google Home in the US market.
- The adoption rate of smart functionality for appliances such as lights, locks, and cookers is uncertain, varies by country, and risks mirroring the disappearance of niche electric tools like carving knives after the 1960s and 1970s.
- A corporate strategy for a unified home system to lock in customers faces a low likelihood of success due to long replacement cycles and the necessity for individual product sales.
- Voice recognition accuracy is expected to improve within the next year or two due to machine learning, addressing issues with misinterpreting similar-sounding words.
- The scalability of voice assistants is limited by the requirement for humans to pre-program answers, meaning systems cannot reliably scale to handle 50,000 or 100,000 unique queries.
- Current voice interfaces function as a "command line" where users must remember available commands rather than serving as a natural language interface, similar to early search models.
- Voice technology may succeed as a universal UI for narrow domains like controlling specific lights or heating in hotels, but general queries requiring pre-built skills remain problematic.
- Voice assistants often fail with complex device controls like thermostats but work reliably for binary commands such as opening or closing drapes in hotel scenarios.
- The ideal product approach involves minimizing on-device intelligence and UI, placing complexity on the network or chip, and using simple controls like a single "go" button.
- Users will likely prefer devices that remove friction, such as kettles with automatic shut-off, mirroring the value of simple, reliable utility over complex interaction.
- Future devices like door locks should operate with the reliability of 2000-year-old technology using advanced chips and computer vision, requiring no screen for user interaction.
- Major electronics companies may accelerate progress by making bold choices to remove legacy components, accepting short-term customer friction for long-term adoption of new standards.
- Embedding Amazon Alexa or Google Assistant runtimes in every device may lack strategic logic for individual device makers and consumers, despite benefiting the platform providers.
- Google and Amazon are currently experimenting with various AI consumer propositions including voice boxes, glasses, and text assistants, with no resolution on the primary product yet.
- Manufacturers cannot rely on an experimental approach for high-stakes products where a single error, such as a voice assistant giving incorrect timing advice, carries severe consequences.
- Unlike search engines which allow manual correction from 10 results, voice commands do not offer a list of interpretations, limiting error recovery.
- While the technology landscape is rich for product development, market changes are not expected to occur as rapidly as suggested by blog posts, urging a focus on solving specific daily problems.