Given the cost of laser diodes is now pennies, how much would it cost to target every window in a city?

You still have to distribute and read those sensors.

I'm imagining drones putting sensor domes with ~100 laser diodes each on top of existing street lights, using in the order of a few watts from a small solar cell and battery to do basic signal processing (most windows will be silent most of the time) and uploading anything possibly interesting over a standard mobile data connection.

Then process the possibly interesting signals with a voice-to-text system similar to Alexa/Siri/Google/Cortana and now there's a searchable text database of everything spoken next to a window in a city.

That would be, what, mid tens to mid hundreds of millions of £ for London? Compared to the annual budget of £3.24 billion?

edit: possibility->possibly

Well, yes. Quite a possibility. In fact you wouldn't need the drones. People putting them there will receive much less attention.

I still can't imagine those globes deciding what is "interesting" and what is not on a solar panel power budget. But at this point it's guaranteed to become viable at some point, and I don't have all the info to conclude it's not viable now.

But when you defend cowardice and the status quo, you don't need arguments. You just need to smirk and say something utterly dumb like the comment about glass windows, and you're mostly golden.