> then obviously bigger businesses like MSG can still do even manual barring 'at scale' (e.g. hire more security guards and distribute lists of barred individuals across them)
Are you sure about that? Image recognition isn't a horizontally scalable task if you're talking about guys with clipboards. And the larger that clipboard, the lower the accuracy gets. Think of it as you have to match each incoming patron against a database. If each security guard has the entire database, they're doing a costly time-intensive, & inaccurate search through the book (security guards are not only paid too little to care about this, but humans in general are going to perform poorly at a task of "does a picture of this person exist in these other 200 images of banned people taken in alternate lighting conditions, clothing, facial hair, etc". And remember, as a business your goal is to let in all the people who came in a timely fashion. Otherwise your business dies / people complain / laws get changed. So to do this, you'd basically have to have a close to 1:1 relationship between person entering and security guard screening. That doesn't scale, even for larger businesses.
Now you could invert the problem. You have the security guard take an image and distribute it to other guards who have their portion of their database that they're combing through and match against a subset of that. Still, that's a heck of a lot of extra man power that scales with the amount of bans you hand out and there's a natural backpressure for your business.
With automation though, you can scale this for pennies. So you don't have to be very discriminant about who you're banning which is a meaningful distinction. This isn't a bad PR approach. You probably can't muster enough political will to amend the laws to force venue owners to only ban people from public venues who themselves are being a public nuisance of some kind (which I think solves the problems a heck of a lot better than protected classes). But you can make it financially infeasible for them to do this by banning the new tech that lets them go after you.
What I meant was that each security guard has a certain subset of the database, so they're only responsible for scanning for those people. Say, instead of having 10 security guards each with a list of all 100 banned people, each security guard has a list of 10 people.
> Now you could invert the problem. You have the security guard take an image and distribute it to other guards who have their portion of their database that they're combing through and match against a subset of that. Still, that's a heck of a lot of extra man power that scales with the amount of bans you hand out and there's a natural backpressure for your business.
I think that's then an open problem of 1) how many people can one security guard reliably track, 2) how many people are on the ban list, 3) is it cheaper to hire that necessary number of security guards for each event than it is to invest in facial recognition tech? 4) How much cheaper or more expensive is it, exactly?
But my main question wasn't about the technical implementation, it was once again about what the actual argument is. Is it that businesses should be allowed to bar people as long as they don't use technology to do so? How would that account for different businesses having wildly different enforcement capabilities?
> Is it that businesses should be allowed to bar people as long as they don't use technology to do so? How would that account for different businesses having wildly different enforcement capabilities?
The issue is that the larger/cheaper the enforcement capabilities, the more likely frivolous bans are to be handed out.
Being able to implement small scale bans is reasonable including a "refuse the right to bar from service for any reason" type clause (which is obviously overridden by any protected groups legislation), but once implementing a permanent ban for large groups is a few clicks in a dashboard there's an issue.
Are you sure about that? Image recognition isn't a horizontally scalable task if you're talking about guys with clipboards. And the larger that clipboard, the lower the accuracy gets. Think of it as you have to match each incoming patron against a database. If each security guard has the entire database, they're doing a costly time-intensive, & inaccurate search through the book (security guards are not only paid too little to care about this, but humans in general are going to perform poorly at a task of "does a picture of this person exist in these other 200 images of banned people taken in alternate lighting conditions, clothing, facial hair, etc". And remember, as a business your goal is to let in all the people who came in a timely fashion. Otherwise your business dies / people complain / laws get changed. So to do this, you'd basically have to have a close to 1:1 relationship between person entering and security guard screening. That doesn't scale, even for larger businesses.
Now you could invert the problem. You have the security guard take an image and distribute it to other guards who have their portion of their database that they're combing through and match against a subset of that. Still, that's a heck of a lot of extra man power that scales with the amount of bans you hand out and there's a natural backpressure for your business.
With automation though, you can scale this for pennies. So you don't have to be very discriminant about who you're banning which is a meaningful distinction. This isn't a bad PR approach. You probably can't muster enough political will to amend the laws to force venue owners to only ban people from public venues who themselves are being a public nuisance of some kind (which I think solves the problems a heck of a lot better than protected classes). But you can make it financially infeasible for them to do this by banning the new tech that lets them go after you.