It's a reasonable idea. Certainly it makes sense that AI companies need training data and people should have a way to sell that data to them. That said, given that the government seems to be moving in a "copyright does not apply to AI" direction, I would not want to putting my eggs in an "AI companies will pay for data" basket.
lurk2 · 6m ago
Private repositories of data will still be valuable due to AI content making public content not suitable for training purposes; but if copyright doesn’t apply the real issue is identifying the content, not paying for it; paying for it might be worse as you’re creating an incentive to upload junk data.
Avipat_ · 19m ago
They are actively paying for it now, and regulation is moving in that direction as well. Regardless, someone's camera roll data, for example, isn't publicly available, meaning it can be sold, unlike someone's Instagram photos.
Avipat_ · 45m ago
We launched Kled 6 weeks ago as a data marketplace for everyday people. Users upload personal content like camera rolls, dashcam footage, homework, POV recordings, and original music, and we pay them for it.
While AI companies scrape public data to train their models, we’re building a platform that compensates users and gives AI labs access to licensed, high-quality data.
The product took off quickly. We did over 20 million impressions on Twitter, more than 10,000 people signed up, petabytes of data were uploaded, and we’ve paid out over $300,000 to users.
Every file uploaded is automatically classified and made searchable so AI companies can instantly license the exact datasets they need.
Over the next five weeks, we’re doubling down on our labeling operations and hiring top AI engineers. We’ve raised over $2 million in venture capital at a $40 million valuation. High pay and high equity. Email avi@kled.ai
No one cares about the aggregated sum you spent, and something tells me that instead of directly saying how much the average user was paid, the PR team chose to divulge the total amount because it sounds better (someone must have pat themselves on the back when they phrased it as "we paid users $300K").
Avipat_ · 17m ago
the money came directly from AI Labs paying for it. not venture or anything else.
lurk2 · 15m ago
That isn’t the issue. If you paid $300,000 to 300,000 users, each user made a dollar.
Avipat_ · 8m ago
We paid $300,000 to 500 users, there's over 10,000 on the waitlist still. Each of them have uploaded over 200GB of data. Some at the 1TB limit.
Wait a minute, if users upload their camera rolls, doesn't that mean that in most cases these also contain footage of __other__ people?
Sorry, but wtf?!?
Avipat_ · 24m ago
Correct but the person who took the pic/vid owns the rights for it legally speaking. This content is already being trained off of, scraped off YouTube, Instagram etc. All Kled does is put the money back into the pockets of the people.
lurk2 · 15m ago
This will never fly in the EU.
Avipat_ · 7m ago
working on it
BolexNOLA · 22m ago
Do they? That’s not an assumption you can make. I’ve likely got plenty of things in my phone I’m not allowed to have just by accident
amelius · 21m ago
Yeah, this sounds like just another sleazy way of a tech company to shove the liability to the user.
gyanchawdhary · 21m ago
Wow, you’ve cracked the mystery of photography .. sometimes people take pictures of other people. What’s next, discovering that mirrors reflect more than just yourself and that’s a valid wtf moment in your world view too …
amelius · 14m ago
The law has limits on what you can do with your camera and photos. Until now, most people were operating within these limits, so there was no problem. In the modern times of big tech, automation, AI, and large scale photo databases, you simply cannot scale up your simple user's rights without running into the law. And imho, rightly so.
Avipat_ · 20m ago
lol yea, not sure what that guy is on
amelius · 19m ago
Read the rest of the comments in this subthread, then maybe you will understand what this is about.
Avipat_ · 7m ago
Can't reply to ur other comment but yea, Latham and Watkins represents us. The best lawyers on the planet, so yea we've talked to lawyers about it lol.
Avipat_ · 15m ago
I've replied to every comment lol, i can assure you this take is braindead
Avipat_ · 2m ago
I've literally never used HN before, a friend suggested I write a post on here for hiring. Lol last thing I care about is self promotion, we're doing just fine.
lurk2 · 9m ago
You are a 21 year old self-promoter and your post was flagged in less than 30 minutes. You don’t have the kind of clout where you can afford to be calling people braindead on an account tied to your LinkedIn profile.
amelius · 12m ago
Have you talked to lawyers?
gyanchawdhary · 12m ago
HN is the UFC of capitalism .. blood, code, sweat and capital. You showed up barefoot with a manifesto about non violence … Wrong arena champ.
BolexNOLA · 23m ago
I’m also curious about this. I’m not having a strong reaction yet because I’m just not quite sure what to make of it, though at first glance it certainly has me going “wait what?”
I hope it’s not just some boiler plate legalese about how you have to make sure you have the rights to everything you upload.
Avipat_ · 21m ago
It’s not just boilerplate. We actively verify uploads using multiple signals: EXIF metadata, timestamps, device signatures, and visual similarity checks. If a file wasn't captured by the uploader’s device or shows signs of being scraped/copied, it gets flagged or rejected.
In short, if it's not your original content, it won’t pass. And we’re constantly tightening this loop as we scale.
slowmotarget · 22m ago
It's funny I had a similar idea but with dashcams, for the automotive industry. You could pay people to install additional cameras to there cars and pay them by the miles driven. I'm sure car manufacturers would love this training data.
Avipat_ · 18m ago
we have around 4000 uber drives that are being onboarded now all uploading their dashcam footage. will see how valubale that data is soon.
While AI companies scrape public data to train their models, we’re building a platform that compensates users and gives AI labs access to licensed, high-quality data.
The product took off quickly. We did over 20 million impressions on Twitter, more than 10,000 people signed up, petabytes of data were uploaded, and we’ve paid out over $300,000 to users.
Every file uploaded is automatically classified and made searchable so AI companies can instantly license the exact datasets they need.
Over the next five weeks, we’re doubling down on our labeling operations and hiring top AI engineers. We’ve raised over $2 million in venture capital at a $40 million valuation. High pay and high equity. Email avi@kled.ai
Sorry, but wtf?!?
I hope it’s not just some boiler plate legalese about how you have to make sure you have the rights to everything you upload.
In short, if it's not your original content, it won’t pass. And we’re constantly tightening this loop as we scale.