Microsoft deletes face recognition database
bbc.co.uk
bbc.co.uk
Meanwhile this[0] was posted on some of the entrances to the common buildings on Microsoft's campus for some time. They seemed to be training an algorithm for "fairness" by taking video of employees that were entering via certain doors. You could "opt out" by choosing another entrance though.
> “I’ll sue you,” the door said as the first screw fell out.
> Joe Chip said, “I’ve never been sued by a door. But I guess I can live through it.”
It's depressing how even when some companies are doing the right thing at their own expense, others delve straight into unrestricted greed with no regard for human lives.
well that has basically always been amazon's m.o.
We'd tried it a few years ago and it was too unreliable, but now it's amazingly accurate.
The elevator pitch is to make it so when the employee uses the device, the app can recognize who they are, and depending on how the organization has configured the app, either automatically clock them in or out, or just use the facial recognition as an authentication factor
Our focus is on construction related companies (CAT Ventures is one of our investors), where we've had demand for an old school feeling time clock experience, but with modernized features. Our app right now is oriented around each employee using a personal phone, but many job sites, especially with heavy equipment, don't want people carrying personal phones around.
Having someone else clock in for your isn't always cheating. For example, leaving early so you can drop by the post office on your way home from work to drop off the company mail.
Or the opposite where my wife works: People will punch in for her because she has to go to an off-site meeting or event.
We have the technology.
I know a factory in Chicago that still has huge time-card punching machine from the late 1800's. It's enormous, and looks like a cast iron wheel of fortune. Very cool.
Somehow the company has managed to function for 150 years without adding facial recognition to it.
Previously, I spent I don't know how much time going "now, how long did I actually work last Monday?" whenever I've tried to figure out whether I've worked enough or not (there was, of course, a target hours/week in my contract when I was employed). Now I know, at the push of the button: I worked eight hours, sixteen minutes and three seconds. OK, then!
If that's all factored in then sure, this could be pretty useful.
A: "Because we don't have anyone to maintain it"
News Media: "Microsoft deletes database".
Does anyone feel like today's news articles, are written by gossip/tabloid writers?
And why shouldn't the database continue to work just because there isn't someone to maintain it? It's not like the maintainer is actually hand-cranking the generator that powers the database server.
And yeah there's definitely a trend with trying to make every news story look like a big dramatic news event with lots of intrigue and "tea". Mostly it's just pretty boring and undramatic stuff.
Also if it's not being used why would you keep it up and running.
finally not sure why you're against someone getting rid of a facial recognition database
The concept of real value was thrown out a long time ago.
> More recently, Microsoft rejected a request from police in California to use its face-spotting systems in body cameras and cars.
Sounds like they are actually starting to get some principles and are standing up for them.
Are there any citations or press releases of Microsoft saying it was deleted?
Microsoft told the FT the database was no longer available, because the person who curated it had now left the company.
People champion pet projects all the time. At my company we joke that every major website UI redesign is just another VP pissing on the web portal fire hydrant.
Without a key person to pitch the value of this project's continued existence it got killed.
Edit: If MS is anything like the BigCo I work at those servers are gonna be wiped and re-imaged after a bunch of automated "these servers were owned by X who has left the company, do you still need them?" emails are ignored by X's boss.
I ran their web site. When my job got outsourced, I got exactly four minutes notice before being shown the door. I was (and still am) the only one with the password for the domain registrar. The people who own the company have too much arrogance and pride to call me to ask for the password so they can give it to the Indian company that replaced me, so after several weeks of the web site not being updated with the current pricing (which changed almost daily), they opened a new domain.
As of last month, my site still sits there as a time capsule with the old pricing, employee bios, and everything from six years ago. I can understand why the domain hasn't changed, because they paid for 10 years in advance. But I don't know why they're still paying the hosting bill.
Nope. No severance. And as I mentioned, almost zero notice. Just "Turn in your badge," and my immediate boss telling me that the project was already outsourced a week earlier.
Something similar happened when I worked at a small town radio station in Wisconsin. This was back when computers were just coming to small business, and people set BIOS passwords and used those round keys to lock their computers. When I got shown the door, nobody asked me for the password or the key to my computer. They just said here's your last check, get out. I heard from one of the DJ's a few weeks later that they ended up chucking out the computer and buying a new one.
The company was run poorly and the passwords to most services were held by random people in the company. He did get the key to the kingdom (the domain name) but he's just too incompetent to realize that with domain ownership he can just set up new email hosting and use password reset on all the services... Instead he keeps making legal threats to get things he trivially can get access to himself.
If they only deleted the training data, but not the ML models generated from them, then you get the worst of both worlds (people still using the models to do things, and no way to validate or improve the fairness of said models by adding or removing labelled training data).
It would be interesting to find out if that were the case, considering that it forms the basis of Azure ML Ops (along with Pipelines), introduced at Build a month ago, with an aim to track assets and create an 'end-to-end' audit trail.
https://azure.microsoft.com/en-gb/blog/take-your-machine-lea...
It's the Asian subset, 4 files, about 300 Gigs.
http://trillionpairs.deepglint.com/overview
addl info
Maybe some of the images in the database were scraped from LinkedIn or GitHub