Owner operators have a long history of declining wages because they are stubborn and anti-union. So if they hate this bill it's probably good for them. Also truck driving is almost certainly going to be the first category of vehicle automated.
Owner operators have a long history of declining wages because they are stubborn and anti-union. So if they hate this bill it's probably good for them. Also truck driving is almost certainly going to be the first category of vehicle automated.
An increasing number of trucking companies are also requiring GPS tracking for contractors, which really pushes things from provider to employee. If nothing else, legislation that pushed back the monitoring of independent contractors that makes them effectively employees would help a lot of people.
As someone who has worked as an independent contractor for an employer that required 8-5 hours in the office, I can tell you that it would make a lot of people's lives easier to have employer control of contractor work conditions better defined by law, especially if that included a clarification that people who are subject to direct management are employees not independent labor.
Owner operators have a long history of declining wages because they are stubborn and anti-union. So if they hate this bill it's probably good for them. Also truck driving is almost certainly going to be the first category of vehicle automated.
An increasing number of trucking companies are also requiring GPS tracking for contractors, which really pushes things from provider to employee. If nothing else, legislation that pushed back the monitoring of independent contractors that makes them effectively employees would help a lot of people.
As someone who has worked as an independent contractor for an employer that required 8-5 hours in the office, I can tell you that it would make a lot of people's lives easier to have employer control of contractor work conditions better defined by law, especially if that included a clarification that people who are subject to direct management are employees not independent labor.
To the best of my recollection, that has been clarified in a few lawsuits. Namely the NLRB v Microsoft back in the early 00s. This is how you got that 18 months in a position with 6 months off before being rehired to do the same exact thing cycle. Funny how instead of simply hiring people tech companies will incur the same onboarding costs over and over again to "save" their bottom line. Which says it's not about money and profit but rather about power. Exactly how much of my education re: business has been outright wrong? I'm going to guess all of it.
All opinions are my own and in no way reflect that of my employer.
+2
Options
HacksawJ. Duggan Esq.Wrestler at LawRegistered Userregular
Owner operators have a long history of declining wages because they are stubborn and anti-union. So if they hate this bill it's probably good for them. Also truck driving is almost certainly going to be the first category of vehicle automated.
An increasing number of trucking companies are also requiring GPS tracking for contractors, which really pushes things from provider to employee. If nothing else, legislation that pushed back the monitoring of independent contractors that makes them effectively employees would help a lot of people.
As someone who has worked as an independent contractor for an employer that required 8-5 hours in the office, I can tell you that it would make a lot of people's lives easier to have employer control of contractor work conditions better defined by law, especially if that included a clarification that people who are subject to direct management are employees not independent labor.
To the best of my recollection, that has been clarified in a few lawsuits. Namely the NLRB v Microsoft back in the early 00s. This is how you got that 18 months in a position with 6 months off before being rehired to do the same exact thing cycle. Funny how instead of simply hiring people tech companies will incur the same onboarding costs over and over again to "save" their bottom line. Which says it's not about money and profit but rather about power. Exactly how much of my education re: business has been outright wrong? I'm going to guess all of it.
That's because most business education in America is bullshit propaganda aimed at guiding the manager class in the ways of exploiting the labor of the working class to pad the pockets of the capital class using buzzwords and the passive voice. It's capitalist equivocation at its best.
Nearly every Revver who spoke with The Verge said they were exposed to graphic or troubling material on multiple occasions with no warning.
“I’ve finished more than one file in tears because listening to someone talk about being abused or assaulted is emotionally taxing, and frankly I have no training or expertise that really helps me cope with it,” one Revver tells The Verge.
Usually, the only indicators a Revver has as to what’s inside a file is its title and the name of the person submitting it, should it be a person or entity they recognize.
If the transcriber chooses to not finish a job after the hour-long grace period, they risk receiving a lowered “commitment ratio score” in addition to giving up any money they would have made. If a transcriber’s standing goes below a certain threshold, their account can be terminated. “Sometimes you’ll get part way through a file and realize, ‘Oh, this involves some child abuse,’” says another Revver. “And if it’s past the unclaimed time, you can’t [abandon] it.”
(Hey, Verge? Don't call them Revvers. I suggest "Transcribers".)
But don't worry, Rev responded by adding a minimum age requirement. (On an unrelated note, before this there was no minimum age requirement to work for Rev.)
As a second bit of Rev news, someone pointed out that Rev lets their transcribers listen to audio you upload. One the one hand, duh. On the other hand, this means Rev is advertising a level of confidentiality they can't deliver on. In a business context, if a company uses Rev, a competitor could totally have someone sign up, watch for company audio files to show up, and choose to transcribe them. It gets worse when you realize Rev is used by police and other legal organizations.
Damn, they cant have 12 year olds transcribe court documents anymore, damn regulation.
How is anyone supose to make money with this much oversight.
I do wonder if using voice to text first, looking for any obvious red flag words or phrases and then giving it over to a transcriber to verify and clean up the text might be a good mix of technology and human work.
But that would be too expensive, better to traumatize contractors working for below minimum wage.
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
He's a shy overambitious dog-catcher on the wrong side of the law. She's an orphaned psychic mercenary with the power to bend men's minds. They fight crime!
Damn, they cant have 12 year olds transcribe court documents anymore, damn regulation.
How is anyone supose to make money with this much oversight.
I do wonder if using voice to text first, looking for any obvious red flag words or phrases and then giving it over to a transcriber to verify and clean up the text might be a good mix of technology and human work.
But that would be too expensive, better to traumatize contractors working for below minimum wage.
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
A lot of the "innovation" of the gig economy can be summed up in two words: fuck regulation.
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
You forget about how much of the AI voice recognition ended up being just paying people to do "the tech" under the table.
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
You forget about how much of the AI voice recognition ended up being just paying people to do "the tech" under the table.
I've used these services and they seem too fast for it to be people.
I'm willing to bet that if the story was "Rev charging higher rates for transcribing abuse victim testimony" this thread's reaction would not be "Oh yeah that's a necessary and proper cost for protecring transcribers" So I'm not sure how you think the system should work.
While racing light mechs, your Urbanmech comes in second place, but only because it ran out of ammo.
0
Options
Gabriel_Pitt(effective against Russian warships)Registered Userregular
I'm willing to bet that if the story was "Rev charging higher rates for transcribing abuse victim testimony" this thread's reaction would not be "Oh yeah that's a necessary and proper cost for protecring transcribers" So I'm not sure how you think the system should work.
If only there was SOME way to have a middle ground, like, ANY sort of content labeling so people actually knew ahead of time what had been submitted, so someone who picks up 'session13.wav' to transcribe doesn't get hit with child rape and abuse out of the blue.
Also the fact that up until now there were no age restrictions on the transcriber.
And that things that should absolutely have a chain of custody in handling and confidentiality, don't.
But if you're just here to play the game, 'invent bad faith hypotheticals' you're the champion, bro.
I'm willing to bet that if the story was "Rev charging higher rates for transcribing abuse victim testimony" this thread's reaction would not be "Oh yeah that's a necessary and proper cost for protecring transcribers" So I'm not sure how you think the system should work.
At a minimum, audio with disturbing content should be flagged, so that transcribers can avoid it or mentally prepare for it as needed. The presence of disturbing content should never come as a surprise.
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
You forget about how much of the AI voice recognition ended up being just paying people to do "the tech" under the table.
I've used these services and they seem too fast for it to be people.
The tech is definitely much better now but speech recognition ran for years powered in large degree by human ears.
A fuckton of AI was and is still that, in fact
That being said it would be ridiculously cheap to use a machine to flag disturbing content before turning it over to a human...in fact those types of flags/categories come out of the box on all call center speech analytics solutions already and have for years!
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
You forget about how much of the AI voice recognition ended up being just paying people to do "the tech" under the table.
I've used these services and they seem too fast for it to be people.
The tech is definitely much better now but speech recognition ran for years powered in large degree by human ears.
A fuckton of AI was and is still that, in fact
That being said it would be ridiculously cheap to use a machine to flag disturbing content before turning it over to a human...in fact those types of flags/categories come out of the box on all call center speech analytics solutions already and have for years!
Oh that exists already? Hmm.. then the choice is spend a few extra cents per file to catagorize it using proven tech or traumatize workers.
Yes it is better to traumatize the workers then lose any amount of money. And if they stop half way through the company can keep their work without paying. Win win!
He's a shy overambitious dog-catcher on the wrong side of the law. She's an orphaned psychic mercenary with the power to bend men's minds. They fight crime!
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
While racing light mechs, your Urbanmech comes in second place, but only because it ran out of ammo.
0
Options
Blackhawk1313Demon Hunter for HireTime RiftRegistered Userregular
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
Oh it might not be perfect, better not do it at all. The bullshit mantra of many a business. Moderation comes to mind.
Damn, they cant have 12 year olds transcribe court documents anymore, damn regulation.
How is anyone supose to make money with this much oversight.
I do wonder if using voice to text first, looking for any obvious red flag words or phrases and then giving it over to a transcriber to verify and clean up the text might be a good mix of technology and human work.
But that would be too expensive, better to traumatize contractors working for below minimum wage.
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
A lot of the "innovation" of the gig economy can be summed up in two words: fuck regulation.
Most frustrating is that they put a nice wrapping around those two words and instead use the other two words: "market disruption" to make it seem like they're actually not being villians.
+8
Options
Gabriel_Pitt(effective against Russian warships)Registered Userregular
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
Oh it might not be perfect, better not do it at all. The bullshit mantra of many a business. Moderation comes to mind.
Again, this seems to be ignoring the simple solution of when submitting files for transcription, the submitter is presented with a few check boxes that identify extreme content, and if transcriber gets a 'clean' file that is actually full of awfulness, they can drop it without penalty.
+10
Options
Blackhawk1313Demon Hunter for HireTime RiftRegistered Userregular
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
Oh it might not be perfect, better not do it at all. The bullshit mantra of many a business. Moderation comes to mind.
Again, this seems to be ignoring the simple solution of when submitting files for transcription, the submitter is presented with a few check boxes that identify extreme content, and if transcriber gets a 'clean' file that is actually full of awfulness, they can drop it without penalty.
That would mean someone having to review the “clean” file and confirm it was bad.
Gabriel_Pitt(effective against Russian warships)Registered Userregular
edited December 2019
Who cares? That's Rev's problem. If they're going to allow not just explicit, but disturbingly violent or otherwise hard to deal with content as part of their service, handling it properly is going to have to be part of their operation.
Gabriel_Pitt on
+8
Options
Blackhawk1313Demon Hunter for HireTime RiftRegistered Userregular
Who cares? That's Rev's problem. If they're going to allow not just explicit, but disturbingly violent or otherwise hard to deal with content as part of their service, handling it properly is going to have to be part of their operation.
Just noting yet another additional thing on it when they can’t even be bothered to try the first thing. Or anything. That they are garbage and should have to do something isn’t even a question and you seem to have taken what I said like I somehow cared about how they feel.
Who cares? That's Rev's problem. If they're going to allow not just explicit, but disturbingly violent or otherwise hard to deal with content as part of their service, handling it properly is going to have to be part of their operation.
Rev can hire "editor supervisors" or whatever the fuck.
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
You have no clue what you’re talking about
Software can already do 85%+ accurate transcription- there are just some cases where you need 100% and thus need a service like Rev
You only need about 10% accuracy to flag sensitive content
This is a simple problem to care for- even without software you could introduce adjustments to your compensation plan and second-level review, as someone else mentioned
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
You have no clue what you’re talking about
Software can already do 85%+ accurate transcription- there are just some cases where you need 100% and thus need a service like Rev
You only need about 10% accuracy to flag sensitive content
This is a simple problem to care for- even without software you could introduce adjustments to your compensation plan and second-level review, as someone else mentioned
Not quite, I have to say. You would need a decently low speech recognition word error rate to be able to have a vague change of recognizing problematic content.
And even with perfect speech recognition, the system you use to classify the content would not be than good, given the current state of the art.
At lot of that is going to be contextual, and the language used would be rather complex and not that easy to learn.
Also, someone would need to read a lot of transcribed horrifying stuff just to be able to train the classifier in the first place... Like, 10,000 of examples, at a minimum, given the various ways things can be horrible.
Understanding the sense of a text is a hard problem for AI. Automatic audio to text transcription is pretty good right now. But developing a system that knows the difference between an article on “The rape of the lock” and an article on rape is very hard.
Understanding the sense of a text is a hard problem for AI. Automatic audio to text transcription is pretty good right now. But developing a system that knows the difference between an article on “The rape of the lock” and an article on rape is very hard.
You could just have something like: Have people submitting disturbing content flag their files as such. Do a computer pass and flag things for human review. Introduce systems so people could report unflagged disturbing content without penalty. Stuff like that.
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
You have no clue what you’re talking about
Software can already do 85%+ accurate transcription- there are just some cases where you need 100% and thus need a service like Rev
You only need about 10% accuracy to flag sensitive content
This is a simple problem to care for- even without software you could introduce adjustments to your compensation plan and second-level review, as someone else mentioned
Not quite, I have to say. You would need a decently low speech recognition word error rate to be able to have a vague change of recognizing problematic content.
And even with perfect speech recognition, the system you use to classify the content would not be than good, given the current state of the art.
At lot of that is going to be contextual, and the language used would be rather complex and not that easy to learn.
Also, someone would need to read a lot of transcribed horrifying stuff just to be able to train the classifier in the first place... Like, 10,000 of examples, at a minimum, given the various ways things can be horrible.
This technology already exists and has worked fine for this purpose for at least a decade
You get these flags out of the box on every speech analytics tool in the market
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
You have no clue what you’re talking about
Software can already do 85%+ accurate transcription- there are just some cases where you need 100% and thus need a service like Rev
You only need about 10% accuracy to flag sensitive content
This is a simple problem to care for- even without software you could introduce adjustments to your compensation plan and second-level review, as someone else mentioned
Not quite, I have to say. You would need a decently low speech recognition word error rate to be able to have a vague change of recognizing problematic content.
And even with perfect speech recognition, the system you use to classify the content would not be than good, given the current state of the art.
At lot of that is going to be contextual, and the language used would be rather complex and not that easy to learn.
Also, someone would need to read a lot of transcribed horrifying stuff just to be able to train the classifier in the first place... Like, 10,000 of examples, at a minimum, given the various ways things can be horrible.
This technology already exists and has worked fine for this purpose for at least a decade
You get these flags out of the box on every speech analytics tool in the market
Given that I'm literally working on that technology, we have very different definitions of "working fine". We still have problems classifying movies reviews as positive or negative...
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
You have no clue what you’re talking about
Software can already do 85%+ accurate transcription- there are just some cases where you need 100% and thus need a service like Rev
You only need about 10% accuracy to flag sensitive content
This is a simple problem to care for- even without software you could introduce adjustments to your compensation plan and second-level review, as someone else mentioned
Not quite, I have to say. You would need a decently low speech recognition word error rate to be able to have a vague change of recognizing problematic content.
And even with perfect speech recognition, the system you use to classify the content would not be than good, given the current state of the art.
At lot of that is going to be contextual, and the language used would be rather complex and not that easy to learn.
Also, someone would need to read a lot of transcribed horrifying stuff just to be able to train the classifier in the first place... Like, 10,000 of examples, at a minimum, given the various ways things can be horrible.
This technology already exists and has worked fine for this purpose for at least a decade
You get these flags out of the box on every speech analytics tool in the market
Given that I'm literally working on that technology, we have very different definitions of "working fine". We still have problems classifying movies reviews as positive or negative...
I could probably help you then as I’ve used this technology to generate $50mm in benefit to my company since 2012!
My experience is that you can get business value for far less precision than the engineers are comfortable with, which is my point with Rev
It seems like if software could already do the transcription then it would be instead of paying more for a person to do it. Your plan would result in a bunch of false positives, a bunch of false negatives, and probably lots of files that simply are too poor quality to work with the software. Who knows if enough useful flagging would actually be produced to make it better instead of worse.
You have no clue what you’re talking about
Software can already do 85%+ accurate transcription- there are just some cases where you need 100% and thus need a service like Rev
You only need about 10% accuracy to flag sensitive content
This is a simple problem to care for- even without software you could introduce adjustments to your compensation plan and second-level review, as someone else mentioned
Not quite, I have to say. You would need a decently low speech recognition word error rate to be able to have a vague change of recognizing problematic content.
And even with perfect speech recognition, the system you use to classify the content would not be than good, given the current state of the art.
At lot of that is going to be contextual, and the language used would be rather complex and not that easy to learn.
Also, someone would need to read a lot of transcribed horrifying stuff just to be able to train the classifier in the first place... Like, 10,000 of examples, at a minimum, given the various ways things can be horrible.
This technology already exists and has worked fine for this purpose for at least a decade
You get these flags out of the box on every speech analytics tool in the market
Given that I'm literally working on that technology, we have very different definitions of "working fine". We still have problems classifying movies reviews as positive or negative...
I could probably help you then as I’ve used this technology to generate $50mm in benefit to my company since 2012!
My experience is that you can get business value for far less precision than the engineers are comfortable with, which is my point with Rev
Given the precision involved, you still need human review. It's not even close to a solved problems, and pretending it is is just a good way to get terrible outcomes while deluding oneself into thinking that things are fine.
0
Options
FencingsaxIt is difficult to get a man to understand, when his salary depends upon his not understandingGNU Terry PratchettRegistered Userregular
edited December 2019
We aren't talking perfect here. We just need to flag them, and have them reviewed by people, and if we have false positives, so what?
We aren't talking perfect here. We just need to flag them, and have them reviewed by people, and if we have false positives, so what?
False negatives are the problems. Given the consequences, you need human review no matter what.
That being said, the solution is to have professional transcribers, with a contract and confidentiality agreement, and all that. This is not a job for crowdworkers, even if they are qualified.
Understanding the sense of a text is a hard problem for AI. Automatic audio to text transcription is pretty good right now. But developing a system that knows the difference between an article on “The rape of the lock” and an article on rape is very hard.
You could just have something like: Have people submitting disturbing content flag their files as such. Do a computer pass and flag things for human review. Introduce systems so people could report unflagged disturbing content without penalty. Stuff like that.
The problem comes down to how to treat the flagged content. Rev very much wants to pretend this isn't an issue because what you do with potentially disturbing content. Do you just label it and no more? So Rev-ers, not being fucking sociopaths, just avoid it so those self labeled ones languish where if you submit it with no mark you get your work done faster. Oh, is there a penalty for not marking it? This is basically the moderation problem is that this content is going to be "I know it when I see it." It all leads to this content should have a higher price tag on it for the higher emotional labor involved but Rev is never gonna do that. Shit, they just cut their general labor rates back in November(?).
Understanding the sense of a text is a hard problem for AI. Automatic audio to text transcription is pretty good right now. But developing a system that knows the difference between an article on “The rape of the lock” and an article on rape is very hard.
You could just have something like: Have people submitting disturbing content flag their files as such. Do a computer pass and flag things for human review. Introduce systems so people could report unflagged disturbing content without penalty. Stuff like that.
The problem comes down to how to treat the flagged content. Rev very much wants to pretend this isn't an issue because what you do with potentially disturbing content. Do you just label it and no more? So Rev-ers, not being fucking sociopaths, just avoid it so those self labeled ones languish where if you submit it with no mark you get your work done faster. Oh, is there a penalty for not marking it? This is basically the moderation problem is that this content is going to be "I know it when I see it." It all leads to this content should have a higher price tag on it for the higher emotional labor involved but Rev is never gonna do that. Shit, they just cut their general labor rates back in November(?).
As long as you flag it, people can take it if they want. If you have trouble getting people to take it, pay your transcribers more to encourage people to do them.
+9
Options
FencingsaxIt is difficult to get a man to understand, when his salary depends upon his not understandingGNU Terry PratchettRegistered Userregular
Understanding the sense of a text is a hard problem for AI. Automatic audio to text transcription is pretty good right now. But developing a system that knows the difference between an article on “The rape of the lock” and an article on rape is very hard.
You could just have something like: Have people submitting disturbing content flag their files as such. Do a computer pass and flag things for human review. Introduce systems so people could report unflagged disturbing content without penalty. Stuff like that.
The problem comes down to how to treat the flagged content. Rev very much wants to pretend this isn't an issue because what you do with potentially disturbing content. Do you just label it and no more? So Rev-ers, not being fucking sociopaths, just avoid it so those self labeled ones languish where if you submit it with no mark you get your work done faster. Oh, is there a penalty for not marking it? This is basically the moderation problem is that this content is going to be "I know it when I see it." It all leads to this content should have a higher price tag on it for the higher emotional labor involved but Rev is never gonna do that. Shit, they just cut their general labor rates back in November(?).
As long as you flag it, people can take it if they want. If you have trouble getting people to take it, pay your transcribers more to encourage people to do them.
Can a recorder pay a transcriber more? I don't know how it works.
Flag it, pay more for the disturbing content work, add some filters so can avoid it or search for it if they want the extra money and got the mental fortitude for it.
And if you submit content that should have a warning without checking the box you can pay the higher rate plus penalty if you want the work done. But that isnt disrupting the market because I suspect that is what normal transcriber work looks like.
The point of a lot of the gig economy is to shift the cost burden elsewhere, normally onto the workers.
He's a shy overambitious dog-catcher on the wrong side of the law. She's an orphaned psychic mercenary with the power to bend men's minds. They fight crime!
Is the argument really that a couple false negatives slipping through the cracks is somehow worse than literally no system at all to tell people it's potentially disturbing content?
Posts
Sorry about going off topic (but WeWork still sucks).
An increasing number of trucking companies are also requiring GPS tracking for contractors, which really pushes things from provider to employee. If nothing else, legislation that pushed back the monitoring of independent contractors that makes them effectively employees would help a lot of people.
As someone who has worked as an independent contractor for an employer that required 8-5 hours in the office, I can tell you that it would make a lot of people's lives easier to have employer control of contractor work conditions better defined by law, especially if that included a clarification that people who are subject to direct management are employees not independent labor.
To the best of my recollection, that has been clarified in a few lawsuits. Namely the NLRB v Microsoft back in the early 00s. This is how you got that 18 months in a position with 6 months off before being rehired to do the same exact thing cycle. Funny how instead of simply hiring people tech companies will incur the same onboarding costs over and over again to "save" their bottom line. Which says it's not about money and profit but rather about power. Exactly how much of my education re: business has been outright wrong? I'm going to guess all of it.
That's because most business education in America is bullshit propaganda aimed at guiding the manager class in the ways of exploiting the labor of the working class to pad the pockets of the capital class using buzzwords and the passive voice. It's capitalist equivocation at its best.
(Hey, Verge? Don't call them Revvers. I suggest "Transcribers".)
But don't worry, Rev responded by adding a minimum age requirement. (On an unrelated note, before this there was no minimum age requirement to work for Rev.)
As a second bit of Rev news, someone pointed out that Rev lets their transcribers listen to audio you upload. One the one hand, duh. On the other hand, this means Rev is advertising a level of confidentiality they can't deliver on. In a business context, if a company uses Rev, a competitor could totally have someone sign up, watch for company audio files to show up, and choose to transcribe them. It gets worse when you realize Rev is used by police and other legal organizations.
How is anyone supose to make money with this much oversight.
I do wonder if using voice to text first, looking for any obvious red flag words or phrases and then giving it over to a transcriber to verify and clean up the text might be a good mix of technology and human work.
But that would be too expensive, better to traumatize contractors working for below minimum wage.
It does amaze me how much of the innovation in these gig jobs are just old things repackaged in an app to disguise the bad parts that made the old system not be allowed or work.
A lot of the "innovation" of the gig economy can be summed up in two words: fuck regulation.
You forget about how much of the AI voice recognition ended up being just paying people to do "the tech" under the table.
I've used these services and they seem too fast for it to be people.
If only there was SOME way to have a middle ground, like, ANY sort of content labeling so people actually knew ahead of time what had been submitted, so someone who picks up 'session13.wav' to transcribe doesn't get hit with child rape and abuse out of the blue.
Also the fact that up until now there were no age restrictions on the transcriber.
And that things that should absolutely have a chain of custody in handling and confidentiality, don't.
But if you're just here to play the game, 'invent bad faith hypotheticals' you're the champion, bro.
At a minimum, audio with disturbing content should be flagged, so that transcribers can avoid it or mentally prepare for it as needed. The presence of disturbing content should never come as a surprise.
The tech is definitely much better now but speech recognition ran for years powered in large degree by human ears.
A fuckton of AI was and is still that, in fact
That being said it would be ridiculously cheap to use a machine to flag disturbing content before turning it over to a human...in fact those types of flags/categories come out of the box on all call center speech analytics solutions already and have for years!
Oh that exists already? Hmm.. then the choice is spend a few extra cents per file to catagorize it using proven tech or traumatize workers.
Yes it is better to traumatize the workers then lose any amount of money. And if they stop half way through the company can keep their work without paying. Win win!
Oh it might not be perfect, better not do it at all. The bullshit mantra of many a business. Moderation comes to mind.
Most frustrating is that they put a nice wrapping around those two words and instead use the other two words: "market disruption" to make it seem like they're actually not being villians.
Again, this seems to be ignoring the simple solution of when submitting files for transcription, the submitter is presented with a few check boxes that identify extreme content, and if transcriber gets a 'clean' file that is actually full of awfulness, they can drop it without penalty.
That would mean someone having to review the “clean” file and confirm it was bad.
Just noting yet another additional thing on it when they can’t even be bothered to try the first thing. Or anything. That they are garbage and should have to do something isn’t even a question and you seem to have taken what I said like I somehow cared about how they feel.
Rev can hire "editor supervisors" or whatever the fuck.
You have no clue what you’re talking about
Software can already do 85%+ accurate transcription- there are just some cases where you need 100% and thus need a service like Rev
You only need about 10% accuracy to flag sensitive content
This is a simple problem to care for- even without software you could introduce adjustments to your compensation plan and second-level review, as someone else mentioned
Not quite, I have to say. You would need a decently low speech recognition word error rate to be able to have a vague change of recognizing problematic content.
And even with perfect speech recognition, the system you use to classify the content would not be than good, given the current state of the art.
At lot of that is going to be contextual, and the language used would be rather complex and not that easy to learn.
Also, someone would need to read a lot of transcribed horrifying stuff just to be able to train the classifier in the first place... Like, 10,000 of examples, at a minimum, given the various ways things can be horrible.
You could just have something like: Have people submitting disturbing content flag their files as such. Do a computer pass and flag things for human review. Introduce systems so people could report unflagged disturbing content without penalty. Stuff like that.
This technology already exists and has worked fine for this purpose for at least a decade
You get these flags out of the box on every speech analytics tool in the market
Given that I'm literally working on that technology, we have very different definitions of "working fine". We still have problems classifying movies reviews as positive or negative...
I could probably help you then as I’ve used this technology to generate $50mm in benefit to my company since 2012!
My experience is that you can get business value for far less precision than the engineers are comfortable with, which is my point with Rev
Given the precision involved, you still need human review. It's not even close to a solved problems, and pretending it is is just a good way to get terrible outcomes while deluding oneself into thinking that things are fine.
False negatives are the problems. Given the consequences, you need human review no matter what.
That being said, the solution is to have professional transcribers, with a contract and confidentiality agreement, and all that. This is not a job for crowdworkers, even if they are qualified.
The problem comes down to how to treat the flagged content. Rev very much wants to pretend this isn't an issue because what you do with potentially disturbing content. Do you just label it and no more? So Rev-ers, not being fucking sociopaths, just avoid it so those self labeled ones languish where if you submit it with no mark you get your work done faster. Oh, is there a penalty for not marking it? This is basically the moderation problem is that this content is going to be "I know it when I see it." It all leads to this content should have a higher price tag on it for the higher emotional labor involved but Rev is never gonna do that. Shit, they just cut their general labor rates back in November(?).
As long as you flag it, people can take it if they want. If you have trouble getting people to take it, pay your transcribers more to encourage people to do them.
Can a recorder pay a transcriber more? I don't know how it works.
The point of a lot of the gig economy is to shift the cost burden elsewhere, normally onto the workers.
Really?