Maybe Voice is Just a Crappy Interface. Or Maybe Tech Firms Are Out of Ideas

A bit of a lighter one today. No betrayal by beloved sports figures or elite politicians. Just Microsoft being dumb.

So Microsoft is going to release a version of Windows Copilot (the brand name for their imitative AI systems) that is expected to be controlled by your voice. They won’t remove the keyboard and mouse functionality, at least not yet, but they ill encourage people to speak their commands to the computer. The goal, apparently, is for you to speak to the computer like you would a person and have the computer do what you intend. This is incredibly dumb, I think.

Put aside whether or not this will work well — we will come back to that. Let’s assume that it does. It is still a bad idea. First, most people who work with computers work in offices, most of which are open plan, meaning not even cubicle walls to muffle the sounds. Even in places with high cubicle walls, the sound carries. Every single office I have worked in is decorated with signs containing some version of “please shut the fuck up we are trying to work here”. Once, that was literally what the sign said. (That firm didn’t have the greatest HR department, it must be noted). Sound is a concentration, and thus a work, killer.

Imagine the hilarity of you telling your machine to do something and your co-worker’s computer doing it as well. Ah, the fun we will have. Even if they can prevent that — and given how often my Alexa and Siri went off randomly based on the TV or radio, I am skeptical — an office filled with people talking to their computer all the time to get basic functions done sounds like a version of hell.

And I doubt this will even work. Voice is not a great interface. To this day, the voice control for texting on my phone/car garbles my texts. When we had an Alexa I always had to enunciate like I was talking to a baby to get it to do what I wanted. Now, I admit that I talk fast (I once had a a Memphis drive through worker tell me that if I wanted any food, I needed to “slow down and talk right”), but isn’t that the point? These interfaces are not one size fits all, they way a keyboard or a mouse generally is (people with accessibility issues excepted). Even if large language models help, they also hallucinate. Not only does the imitative AI system have to understand you, it has to be able to transfer that understanding to a set of actions, something that even text based imitative AI systems have had trouble doing. They will not work well for every language on top of all their other issues. The models are only as good as the training set and many models do not have a great range of smaller languages to train on. Lord help the Icelanders who have to use these things.

I think, generally, that consumer computer firms are all out of ideas. They are pushing imitative AI because they have nothing as transformative as the personal computer or the iPhone on tap. That lack of imagination then means that they try and shove imitative AI into everything, including trying to generalize interfaces, like voice, that have limited use cases. Basically, they have plucked all the low hanging fruit and haven’t ben able to climb the tree, to butcher a metaphor. And now they are left trying to sell us a world where we all scream at our machines and hope they do what we want.

It’s hard to come up with new ideas when the obvious one shave been taken. But I truly do with they would try harder. I don’t need my life to look like a bad Saturday Night Live sketch.