Hacker News Reader

Comment by forensic | Hacker News Reader

Am I correct in thinking his ideas on sniffing machines are totally ludicrous?

Are we not basically talking about a chemical-based air tester that would have to sort through ridiculously complex masses of data trying to pick out which subtle combination of particles applies to a perp?

The sense of smell of a dog is extraordinarily complex and makes full use of a brain. I don't see this being duplicated by computers anytime soon. We'll have full-on walking and talking robots before then... and those robots won't have a sense of smell for a long time.

Or maybe I'm just ignant.

Seems to me it would be cheaper to build a kill-bot that defends all your possessions from intruders than to make a device that can monitor the subtle smells in a room.

View on HN

No, I don't think so. Airports are already using a similar technology to detect explosive residue. Basically it's a small telephone booth sized box that blasts you with air and sucks it back up to examine. I'm not sure exactly how it works but it seems possible it could scale to detect other types of particles. At a crime scene you could possibly detect unique body fingerprints based on of everything from perspiration on hair, skin particles, clothing particles, cologne, shampoo type, deodorant type, etc) So if the police can get this scan quickly they could go out and round up possible suspects and compare. Unfortunately I think this points towards more brute force policing where innocent people are subjected to "body scans" simply because they vaguely match a description or were in the wrong place at the wrong time.

The machines are airports detect one "kind" of thing, bomb, etc.

He's talking about detecting a unique person out of many - that's much much harder.

You make my point for me. Look at how much effort and expense they have to go to just for those bomb sniffing machines - and they basically can track only a small handful of very specific, very rarely found chemicals.

Implementing this to sort through things like deodorants and skin flakes is way more complex. You are dealing with organic molecules, for starters. And there are billions of different combinations of airborn bacteria and viruses and stuff that make up the smell of a person.

I think his timeline may be too short, but I don't think it is that far off - 20 years, maybe.

I would certainly agree that smell-recognition may be more difficult than image-recognition. However, there's been a lot of progress on that front, more than I expected just a few years ago.

Reply on news.ycombinator.com