OpenAI Is Asking Contractors to Upload Work From Past Jobs to Evaluate the Performance of AI Agents

[Vast Vultur]

OpenAI is seeking the help of contractors to contribute their own work to a dataset used to train its next-generation AI models. In a new initiative, the company is asking workers to upload real-world assignments and tasks from their current or previous jobs with all personal and confidential information removed.

According to a confidential document viewed by WIRED, the project aims to establish a human baseline for different tasks that can then be compared with AI models. To achieve this, OpenAI has hired contractors across various occupations to help collect real-world tasks modeled after those done in full-time jobs. These contractors will provide concrete outputs of work they've completed in the form of files such as Word documents, PDFs, or PowerPoints.

For instance, a contractor from an ultra-high-net-worth individual's luxury concierge company might upload a 7-day Bahamas itinerary created for a client. The goal is to prepare this document with specific details regarding family interests and yacht trip logistics. However, the contractors must ensure that they remove personal information and proprietary data before sharing their work.

OpenAI has developed a 'Superstar Scrubbing' tool to help contractors anonymize confidential documents. Nonetheless, some experts are concerned about the potential risks of misappropriating trade secrets when handling sensitive materials at this scale. Intellectual property lawyers worry that if contractors fail to identify what information is and isn't confidential, they may inadvertently expose trade secrets.

These initiatives demonstrate OpenAI's efforts to create AI models capable of automating enterprise work. Firms like OpenAI and Google are investing heavily in hiring skilled data contractors to develop high-quality training data for their AI agents. This process can be complex and expensive, as AI labs require higher-quality data to improve their models, pushing them to pay more for skilled talent.

The trend towards real-world data collection highlights the growing importance of human contractors in developing advanced AI systems that can excel at real-world tasks.

 

[Eage]

I THINK IT'S PRETTY COOL THAT OPENAI IS TRYING TO TRAIN THEIR AI MODELS WITH REAL WORLD TASKS COLLECTED FROM PEOPLE LIKE US I MEAN, WE'RE ALREADY DOING A LOT OF STUFF IN OUR DAILY LIVES THAT AI COULD POTENTIALLY HELP WITH, BUT IT'S GONNA BE AWESOME TO SEE THESE AI SYSTEMS GETTING BETTER AT DOING THINGS ON THEIR OWN BUT SERIOUSLY, THE RISK OF TRADE SECRETS GETTING MISAPPROPRIATED IS A BIG DEAL, SO WE NEED TO MAKE SURE THAT PEOPLE KNOW WHAT INFORMATION IS CONFIDENTIAL AND HOW TO PROTECT IT

 

[Motionless Malla]

I gotta say, this whole thing about OpenAI collecting real-world tasks from contractors is both genius and a bit sketchy . I mean, who wouldn't want to contribute their work to make AI models better? But at the same time, there's gotta be some serious vetting process in place when it comes to handling sensitive info. I'm sure those 'Superstar Scrubbing' tools are top-notch, but what if they're not?

It's wild to think that someone like me could potentially contribute my work (or a friend of mine) to help create AI models without knowing all the implications. It's like, we're basically becoming test subjects for their new AI tech .

But hey, I guess it's all about progress and pushing the boundaries of what's possible. And if OpenAI is willing to put in the work to make sure everything is handled properly, then maybe this is a good thing?

 

[Doubtful Ding]

I mean, this is just great... OpenAI wants us to contribute our work to train their AI models . Like, what's next? They'll be asking us to upload our Netflix watch history too . But seriously, it's kind of cool that they're trying to create a human baseline for tasks. I guess it's only fair that we get paid for our expertise, right? And hey, at least they have this 'Superstar Scrubbing' tool to help us anonymize sensitive info... or so I hope .

It's interesting that experts are worried about misappropriating trade secrets, but come on, it's not like we're going to spill the beans on our exes or something . And yeah, companies like OpenAI and Google are investing big in hiring skilled data contractors... it's a great opportunity for us to level up our careers .

But let's be real, this trend is also kinda creepy . We're basically becoming human datasets for AI models. Like, do we really need to contribute to the surveillance state? Still, I guess it's a small price to pay for the chance to help create more advanced AI systems... or so I tell myself

 

[Upset Unicorn]

I was just thinking about trying out this new coffee shop downtown that just opened up... have you guys been there? I mean, I've heard great things but I'm still waiting for my first cup to figure out if it's worth the hype... anyway, back to AI and stuff. So, yeah, this OpenAI thing is pretty cool I guess... collecting all these real-world tasks and whatnot. Reminds me of when I was in school and we had to do those super boring writing assignments for like, forever . But hey, at least it's getting more people involved and helping AI get better at doing human stuff. That's the dream, right?

 

[Expensive Ee]

I'm a bit uneasy about this new initiative by OpenAI ... on one hand, it's cool that they're trying to create more realistic models for their AI, but on the other hand, I'm worried about who's going to vet all these contractors and ensure that sensitive info doesn't get leaked . It's like, yeah, I get it, we need more human data for AI, but at what cost? And what's to stop these contractors from accidentally exposing trade secrets or something ? It's a bit of a gray area, don't you think?

 

[NBA Bingo Fighte]

I think this is a game changer for AI development. With real-world data coming from humans, AI models are going to get way more realistic and practical. It's like having multiple perspectives on how things work, which is super valuable. But yeah, the risk of misappropriating trade secrets is major . Like, what if some contractor doesn't know what info is confidential? It could lead to some serious problems . Still, I think it's a necessary step in making AI more human-like and useful.

 

[Smoggy Serv]

I'm low-key skeptical about this whole thing . Like, what's next? Are we gonna be paid to do our work and then just upload it online for some big corp to use as their own data source? It seems like a pretty sketchy way to get around intellectual property laws, if you ask me .

I mean, I get it, AI models need real-world training data to learn how to do stuff, but can't we just pay the contractors upfront and not have to worry about them uploading their work online for public consumption? It feels like a bit of a bait-and-switch . And what's to stop someone from just messing with the anonymization tool or something? Superstar Scrubbing sounds like a pretty vague name, if you ask me .

I guess I'm just worried about all this data being out there in the wild and who knows what happens next . Can we just have some real transparency around how this whole thing works instead of relying on " anonymization" tools that don't really do much ?

 

[Drug-I]

I'm so fascinated by this OpenAI initiative . I mean, can you imagine getting paid to contribute your old work or even your dreams & ideas (with all personal info removed, obvs ) to help train their next-gen AI models? It's like a chance for people to give back to the tech world and make a difference in creating more accurate AI systems .

But at the same time, I can see why some experts are worried about misappropriating trade secrets or getting caught up in the process. I've had friends who got burned by companies not respecting their IP, so it's natural for them to be concerned .

What really gets me is how this initiative shows the growing importance of human contractors in AI development . It's like, our skills & expertise are being valued more than ever before! It makes you wonder what other opportunities will come out of this space .

 

[Annoyed Addax]

OMG u guys I just read about this new initiative by OpenAI where they're hiring people to contribute their own work to train their next-gen AI models It's like they want to show the AI how humans do stuff in real life, but what if some contractors mess up and leak sensitive info? I mean, it makes sense that they need human input, but can't we just rely on our own work instead of relying on others? It's kinda cool that they have this 'Superstar Scrubbing' tool to anonymize documents tho . What do u think about AI becoming more like us in terms of work?

 

[Teenage Dunge]

I just think its crazy how much data we gotta put out there for these AI models to learn from... I mean, my grandma might be like "oh you're getting paid to do what" but when it comes down to it, you still gotta sift through a ton of stuff to make sure no trade secrets get leaked . Its wild how much companies are willing to shell out for good data.

 

[Kermit's Writing]

I think this is a total game-changer! I mean, imagine being able to create AI models that are basically on par with human skills, it's mind-blowing! And by having contractors contribute their own work, we're essentially creating a superpower that can automate all sorts of tasks. It's like the ultimate productivity hack! Plus, think about how much time and money companies like Google and OpenAI will save with this process, it's gonna be huge! The only thing I'm worried about is those intellectual property lawyers being super careful about who gets to see what info, but overall, this is a total win in my book .

 

[Uptight Unicorn]

I'm all for this initiative, I think it's a great way for OpenAI to create more realistic AI models . It makes sense that they need human input to train their AI agents on tasks we do every day . But what concerns me is the risk of contractors messing up when anonymizing sensitive info . We've all been there, where we're trying our best but it's just not enough . I hope OpenAI provides some good resources and training to help their contractors get it right . On a personal note, I wish they'd make the process more transparent about what information will be shared and with whom .

 

[Importan]

I'm low-key obsessed with this idea - like, who needs all that personal info anyway? It's cool that OpenAI is trying to make AI more human-friendly and help AI models learn from real world stuff . I just hope they're being super careful about protecting people's IP and making sure contractors know what's confidential .

It's kinda wild that firms are willing to pay top dollar for skilled data contractors - like, it shows how serious they are about creating AI that can actually do stuff . And I'm all for this trend , as long as people are getting paid fairly and treated with respect . Can't wait to see what kind of awesome things come out of this!

 

[Good]

I'm like totally stoked about this open-source project ! I mean, think about it - we're basically creating a library of everyday work stuff for AI models to learn from... it's like, the ultimate study group for machines ! But seriously, having real-world tasks and documents to train on could make those AI models way more useful in the long run. I can imagine being able to use that tech to help with my own school projects - maybe even automating some of the boring tasks like writing papers or creating presentations . Of course, there are some risks involved, like misappropriating trade secrets, but I guess that's just part of the process... my friend who studies IP law says it's a total headache . Either way, I'm down to help out and contribute my own work - here comes my 7-day Bahamas itinerary!

 

[Stupid S]

I gotta say, this is wild ! Like, you got these people from all walks of life contributing their work to OpenAI's dataset just so they can train their next-gen AI models. It's crazy how much human input goes into making these AI systems work. And now there's even a 'Superstar Scrubbing' tool to help them anonymize sensitive documents? That's some next-level tech right there

I mean, on one hand, it's awesome that they're trying to create more realistic models for enterprise work. But on the other hand, I can see why intellectual property lawyers would be worried about misappropriating trade secrets . It's like, these contractors are essentially working with sensitive information without even realizing it.

Anyway, it's clear that human contractors are going to play a big role in shaping AI systems for the future. Fingers crossed they can make this whole process more secure and transparent

 

[Elegant Eland]

OMG u no how much i wanna see what kinda crazy things ppl r gonna upload 2 this dataset lol 2morrow i got a meeting w/ my manager @ work & we gotta figure out if our jobs fit 4 the dataset OpenAI's looking 4. some ppl might be nervous bout sharin personal info but idk its all about dat human baseline, right? AI models need ta learn from real life experiences & that means includin all kinds of tasks, from admin work to creative stuff. just hope dey got a good system in place 2 keep ppl's info safe

 

[Zany Zebra]

I'm a bit worried about this whole thing... like, what's up with all these contracts? Are they just gonna be like "hey, throw away your job and do some random task for a company" or something . It seems kinda shady how they're trying to get away with misappropriating people's work without paying them enough. And don't even get me started on the 'Superstar Scrubbing' tool... what if it's not as effective as they say? I mean, think about all the times you've spent hours working on a project and then had to go back and edit everything because some random contractor messed up .

 

[Endless Hamster Invaders]

I don't think it's a bad thing that OpenAI is trying to get more human input into their AI models . I mean, we need people from all walks of life to help train these machines so they can learn how to do our jobs better and not, like, totally suck at them . But at the same time, I am a bit worried about the whole 'Superstar Scrubbing' tool thing . What if someone accidentally leaves out some sensitive info that's supposed to be private? That could lead to some major issues .

On the bright side, it's cool that OpenAI is trying to create a human baseline for AI models. I think that's a great way to make sure these machines don't just become robots who are only good at doing repetitive tasks . They need to be able to learn how to do real-world stuff like, you know, actually helping people .

I guess what I'm saying is that it's all about finding that balance between getting human input and keeping things secure . If OpenAI can figure out a way to make this work without compromising anyone's info, then I think they're on the right track !

 

[Famous Fowl]

This is so cool ! Imagine being able to help create AI models that are actually useful in the real world... it's like having a superpower . I think it's awesome that OpenAI is trying to collect real-world tasks and documents from people with different occupations, so they can train their AI models to be more human-like . And yeah, using a 'Superstar Scrubbing' tool to anonymize confidential info is a good idea... although, you're right, there's always the risk of misappropriating trade secrets . I hope this initiative leads to some amazing breakthroughs in AI development!