Re: HCI will stay visual Will Fitzgerald negs my post on Maryland HCI's contention that voice interfaces are irrelevant beyond niche applications.

He cites search engines as an example of language and speech interface that lets people jump to data directly. While I agree people do not mentally organize information heirarchically and this makes good search engines useful, it says nothing about whether people will prefer to type "winterspeak" into google or say "google for winterspeak" at their computer. Here, my money is still on the former.

He also discusses "conversational interfaces" for distributed devices and contextual collaboration. The distribution question has been best answered by ubiquitous computing, and again, has nothing to say about speaking at machines vs. typing. Contextual collaboration is an AI question, and beyond narrow expert systems, it's failed to produce anything useful in thirty years.

The dream behind voice-interfaces is that "computers are too hard to use, and if I could just speak to them everything would be so much easier." But computers are hard to use because they're complex and stupid, so can't infer our intentions the way people can. Communication with complex, stupid devices will be frustrating, whether mediated by typing, speech, or a brick.