I am working on a solution. Without giving it all away, it’s an app for phones that can record audio and video and recognize a fire using ML and AI. It alerts nearby emergency responders based on GPS from the phone.
Initially this service will be ad-supported. When you launch the app you see a short video ad before the app will record the fire. With such a sharply defined demographic the ad impressions should get a good price.
That’s one way to think about it, but smoke detectors aren’t programmed in Rust, Typesscript, React, don’t run in a K8S cluster, aren’t serverless, so I think I have some technical advantages.
The app can either call 911 and report a fire, or you can hold it next to an Alexa/Siri/Google home hub or smart speaker, and some toaster ovens, and it will tell the connected device to call the fire department.
>but smoke detectors aren’t programmed in Rust, Typesscript, React, don’t run in a K8S cluster, aren’t serverless, so I think I have some technical advantages.
Smoke detectors don't require a subscription fee, or for me to sit through advertising. Smoke detectors won't stop working on a bad connection or a botched update. Smoke detectors won't spy on me and sell my data.
Is your app actually, testably, provably better at detecting fires than smoke detectors? Will you have the endorsement of fire departments? Will my homeowners' insurance be affected by using this app?
The ad is a waste of time, I think. You will have to avoid wasting time, in the case of fire. (These ads are problematic for other reasons too, but here is one situation too.)
I'm not surprised nobody has tackled this problem. The result of your "app" breaking is someone who needed help with a fire has their entire home / structure burn down.
I sure as hell don't want a more complicated system to tell me if my house is about to burn down or if I'm about to asphyxiate in my sleep.
The Nest smoke-detector has a horrendous number of examples of exactly this... especially people getting annoyed with false positives because it's over-complicated and then turning off the detector and dying as a result when there was later an actual fire. [0]
Do you think it would be acceptable for a fire department to ignore an emergency call because some AI algorithm tells them it's 92% likely to be false? Who would be responsible if the algorithm made a mistake and people died?
Not necessarily what OP is proposing. It could be a group if people fielding calls to the department who rapidly investigate all calls, and get the fire department there for real fires.
They could also provide some first responder care... maybe this is already what some fire departments do?
To investigate whether a fire call is real, wouldn't you need to be a trained firefighter? For example, you may have to enter a building that's possibly on fire to see if there's a fire. And a small fire isn't necessarily obvious by just looking at the building from the outside.
Also, if someone has to investigate before the fire department was dispatched, you'd lose a lot of valuable time, allowing the fire to spread. It could be a win if there was no fire, but a huge loss (burned buildings, dead people) if there actually was a fire.
Yeah I was playing devils advocate I guess, I do agree with you. The best equipped people to investigate the false alarms is probably the fire department itself, and they are already the ones doing the investigating.
Improve fire detection and reporting systems. Possibly improve in-home and in-business fire suppression systems, or even have them. I live in a fire prone area and none of the houses we looked at had fire suppression systems built-in, definitely something I'll be looking into if we have a chance to build.
But this is definitely an area where false positives are much less of a concern than false negatives. We'd rather have fire and rescue sent to non-incidents than not sent to real incidents.
At my work folks always set off the alarm with the microwave. Usually popcorn. The FD said by law they have to show up and check out the building each time.
Yeah I think regulation is the way. When I lived in Australia, the law and setup was this:
- The in unit smoke detectors can be triggered to warn occupants. No emergency services are called. If you have a cooking mishap or any kind of smoke, it will trigger, but just to wake the people in the unit up or annoy your neighbors.
- There are also smoke detectors in the hallway. Once this is triggered (presumably real fire), emergency services are notified and on their way. The building alarm will also sound. If it is found to be a false alarm, the tenant who triggered it is fined.
I think it's really reasonable. Small mishaps don't need to have a big response, but tenants do need to be notified. Once it's big enough (smoke in the hallway), then a response is justified.
> The in unit smoke detectors can be triggered to warn occupants
This is specifically because the National Construction Code (NCC) requires multiple occupancy buildings to have certain fire safety infrastructure that generally includes a 2 hour fire separation between different occupancy units. This gives a buffer to all other occupants.
> There are also smoke detectors in the hallway.
If a detector is triggered in a common area this has the potential to affect the escape routes of multiple occupants so it triggers an immediate fire alarm.
Australian fire safety rules are some of the best in the world.
God I hope no one on HN ever holds a public office. "No more task forces or comittees or councils. All government reforms will be done via startups from here on out."
Potentially, yes. If fire fighters are busy responding to false alarms, they may be delayed if a real fire happens at the same time. This can result in loss of lives and property.
Initially this service will be ad-supported. When you launch the app you see a short video ad before the app will record the fire. With such a sharply defined demographic the ad impressions should get a good price.