Architecting the Future: Building Hardware for an AI-Native World | TechCrunch Disrupt 2024

Introduction and Panel Discussion

• The panel discussion is moderated by Lee, VP of AI products for HP, who aims to explore the future of work and the role of AI in transforming it (24s).
• The panel consists of Carl Pei, founder and CEO of Nothing, Mark Solomon, prior head of product and design for Pebble, and Jerry Yang, founder and CEO of Brain (1m4s).
• The audience is polled to gauge their involvement in AI, hardware development, and Kickstarter launches (1m26s).
• The discussion begins with the topic of how humans will interact with AI in the future, with Carl Pei suggesting that it should feel natural and seamless, similar to human-to-human interaction (2m18s).

AI Interaction and User Experience

• Carl Pei believes that in the short term, interacting with AI will involve more seamless and personalized devices, while in the long term, it will feel very natural (2m21s).
• Jerry Yang shares his thoughts on how AI will change the experience, citing his company's IOS app called Natural, which features generative interfaces (3m17s).
• Jerry Yang notes that AI will make users feel free, as current tools are passive and pre-built, whereas AI can adapt to changing needs and create new experiences (3m32s).
• The discussion highlights the potential for AI to revolutionize the way humans interact with technology and create more personalized and automated experiences (3m6s).
• Tomorrow's AI-native world will have design decisions, data decisions, and interaction decisions made on the fly, allowing computers to understand humans better and adapt to individuality, resulting in a sense of freedom (4m29s).
• This freedom is similar to a generative UI, where computers can make decisions and adapt to individuality, and use cases that designers and engineers haven't thought about (4m56s).
• The key to making these technologies usable is to think about the medium through which they are shipped, making them natural and user-friendly, which is both a technological and design problem (5m39s).
• The user experience will evolve with AI, and it's interesting to think about how humans will interact with AI, devices, and other humans, as well as AI to AI interactions (6m28s).
• The future of interaction may involve more natural interfaces, such as cybernetics and robotics, which will reduce the learning cost involved in computing by making technology similar to human experience (7m3s).

The Rise of Agents and AI-Native Operating Systems

• Agents are a hot topic, but it's unclear if every brand will have its own agent, and distribution may be a challenge, as people want simplicity in their lives and don't want to learn multiple interfaces (7m37s).
• The idea of every brand having its own agent may not work, as people prefer simplicity and don't want to interface with multiple different interfaces (8m4s).
• The proliferation of agents is expected, with a few potentially winning, and these agents are considered powerful due to their ability to use tools, evaluate their own work, and work with each other (8m16s).
• Modern operating systems have features like multitasking, which allows users to pull up multiple windows at the same time, but this is taken for granted, and the world of agents is about designing interfaces for multiple autonomous agents (8m48s).
• The analogy of cars is used to describe the difference between the current digital world and the world of agents, where instead of designing the steering wheel, the focus is on designing a map or operating on the destination of life (9m11s).
• An AI-native operating system is being developed and will be shipped with major OEMs next year, which will bring a seamless experience and make it more natural to interact with agents (10m4s).

Building AI-Native Products and Hardware

• To build products that bring the AI-native experience to life, it's essential to identify the user market, problem statement, and type of model to use, such as LLM or LVM, and think about the chemistry of the agentic layer (10m43s).
• The concept of agent overload is discussed, where the number of agents can become overwhelming, and it's crucial to vet the agents that will have meaning and solve problems (11m31s).
• The essential components for a truly native generative AI experience include identifying the user market, problem statement, and type of model to use, and then layering in compute and inputs (12m0s).
• A helpful approach is to start with the bare essentials and add only what is necessary, which is useful for items like Gen meets wearables (12m5s).
• When designing AI-native hardware, it's essential to start from the core and solve the general problem, rather than focusing on monetization, which is a separate issue (12m14s).
• Hardware development takes real time and involves significant product-market fit risk, making it challenging for startups to invest resources in innovation (12m41s).
• Instead of investing in new form factors, startups should focus on big categories like smartphones, wireless earbuds, and smartwatches, which have large markets and established volumes (13m18s).
• The goal is to connect existing products and make them more intelligent and serving to consumers, rather than taking the risk of developing new hardware form factors (13m43s).

The Future of User Interface and AI-Native Design

• The current trend is towards seamless interactions, with capabilities running in the background, and the ability to have conversations with foundation models (13m59s).
• The touchscreen is unlikely to go away anytime soon, and the future of user interface will likely be a combination of natural interface and touchscreen (14m14s).
• AI-native design should be centered around natural interactions, with a focus on thoughtful design and human ergonomics, similar to Apple's approach with the iPhone and touchscreen (14m34s).

The Current State and Future Potential of AI

• The current era of AI is still in its early stages, and there is a need to develop a scalable design architecture and technology platform that is reliable and accessible to everyone (15m46s).
• The AI-native world is envisioned as a future where intelligence serves every individual, but currently, it feels like the "fart app stage" of AI-native, where cool little applets are being built rather than fundamentally AI-based things (16m16s).
• As AI-native technology advances, it is expected to unlock superpowers that will revolutionize various aspects of life, such as productivity, creativity, and human augmentation (16m50s).
• Jerry envisions a future where the compression of time will be experienced, allowing for years of work to be done in months or weeks, and the ability to transcend time and linearity of perception (17m5s).
• Mark is excited about the potential of AI-native technology to unlock productivity and enable individuals to do more and better, especially in high-stakes situations such as rocket maintenance or medical emergencies (17m49s).
• Carl believes that the modern smartphone has failed to live up to its potential as a "bicycle for the mind," instead becoming a source of distraction and addiction, and hopes that new AI-native technology can help rebuild a tool that truly empowers creativity (19m27s).
• Humanity's superpower is its ability to find and solve problems in novel ways, and technology products should help humans become more creative (20m25s).
• In an AI-native future, devices are imagined to do more for humans, know more about them, and be more seamless, learning how to interpret humans rather than the other way around (20m39s).

Closing Remarks

• The future of AI-native is described as "nothing" by Carl, followed by laughter, and Mark describes it as "optimistic", "natural", and "awesome" (21m7s).