I doubt documents and websites will go away. Laying out information in text and tables is an extremely efficient way to communicate information quickly. Even th

I doubt documents and websites will go away. Laying out information in text and tables is an extremely efficient way to communicate information quickly. Even the most poorly designed storefront is densely packed with information. In contrast, a conversation with an AI via text, voice, AR, or VR just isn't as dense. It's a very inefficient way to communicate mundane things. If I want to buy some socks, I can hop on Amazon, search for "socks", and find literally tens of thousands of items for sale. I can quickly refine that search with filters or just browse to find what looks good to me. I can find pretty much exactly what I want in less than 30 seconds. Now imagine that same task (buy some socks) through an AI 'virtual assistant' interface. "Hi Alexa, I'd like some socks." "I have over 60,000 socks available to purchase. I recommend 'Dickies Men's Dri-tech Moisture Control Socks Multipack' for $12.99 with free Prime same-day delivery." "Tell me about the socks you recommended." "These socks are currently available in 4 different sizes, which fit as expected, and 60 different colors. They're made of 64% cotton, 33% polyester, 2% spandex, and 1% nylon. They feature soft breathable moisture control fibers, ventilation channels to enhance air flow, arch compression support, and durable reinforced heels and toes." "Are they low cut?" "No, these socks are only available in crew cut. Would you like to search for low cut socks instead?" And so on... You could certainly get to what you want faster than this example, but communicating all of the mundane product information through a conversational interface is a bottleneck. We'll certainly see Ai-driven avatars for some things. They'll be salespeople on car websites and customer service reps on other sites, but regardless of the role they play, they'll supplement the document based interfaces that we already have rather than completely replace them. A good interface is one that makes the user comfortable and enables them to get things done quickly. In terms of getting things done quickly, AI avatars can't really compete with just dumping information on the user. I can look at a list of socks and filter out the ones I don't want much faster than I can communicate to a salesperson what kind of socks I'm looking for. And that's fine. Let the avatar do what it's good at and let the document do what it's good at so you get the best from both.