Am I missing something other commenters are seeing about this not being an ad? The domain is on Burla, which hosted the compute needed for this. There's a giant airbnb x burla logo at the top. People are saying there's a lawsuit pending, it's against guidelines, what's the point, etc..
It's content marketing plain and simple for Burla towards people that view this site. It was highly likely done by employees at both Burla and AirBNB together as a joint project.
One of the Burla founders here. Not a joint project with Airbnb. I’ve been experimenting with giving agents access to Burla clusters and letting them run with analysis ideas I find interesting. This was one of the results.
The branding is a bit much, fair call, but the intent here was just to explore what these agents can actually build when you give them access to large amounts of compute.
This post is a bit lighter on that disclosure than I'd like (and isn't as obvious as a Show HN would be) but I feel I missing some big portion of the backstory to this comment?
I'm struggling a bit with how the 'funniest' ranked reviews are genuine descriptions of people's miserable (and sometimes unsafe) experiences. Where's the funny?
As an experitisement, I guess it gets the name out there but not in any way I'd want for my business.
personally I find those experiences really funny especially in my life. looking back I think most people find humor in it, i could be wrong? I don't think so though
Sure but it's not your life, right? This is other people's misfortunes, and these reviews weren't written to convey their entertainment at an old story.
I found one in Istanbul [0] (which now 404s) that somewhat fits the label and looks like it could have been a set on The Wire, but most of the "drug den" ones are just cramped, taken by someone who doesn't know how to take pictures and doesn't care to learn (blurry, bad lighting, noisy, poor staging), or both.
Most of the bad TV placement ones are also boring because they're just over a fireplace. Technically correct, but not noteworthy. However, I did find one that was truly spectacular [1] (still live for now) and left me with more questions than answers.
I feel like floor mattresses, trash, and peeling paint were also at play. They're all sort of unsafe rooms people wouldn't want to go to unless they felt like they had to (i.e. doing drugs)
>Everything was parallelized on Burla, on a single dynamic cluster that scaled to ~1.7K CPU workers for photo download and CLIP, with 20 A100 GPUs running embedding clusters in parallel on the same cluster.
That's a lot of budget - would have been nice if they'd made an actual donation to the project, instead of pounding the project's servers and bandwidth when there are much better ways to interact with the data.
Totally fair callout. I should’ve been more careful here and leaned on the provided datasets / bulk access instead of pulling things at scale. That’s on me.
I’ll make a donation to support the project regardless. Appreciate you raising it.
This seems like an advertisement for an open source package
>Scale Python across 1,000 CPUs or GPUs in 1 second.
Burla is a high-performance parallel processing library with an extremely fast developer experience. Scale batch processing, vector embeddings, inference, or build pipelines with dynamic hardware.
Edit: Author comment was flagged dead. They work at burla which is a managed cloud service for parallelizing python
These are amazing! Some are probably offensive, because I saw a cozy, if kitschy, British den labeled as "did-someone-just-leave" vibes which... unfair.
What a waste of energy (money/resources)... Scraping and AI-scanning 2 million photos to identify animals in the advertisement pictures? What's the point.
As an exercise a sample of 1000 photos would've been enough. As a database, knowing a listing has a cat in the picture or a funny review doesn't offer any real value.
I wonder what the footprint is of such an exercise.
I dunno there are literally 100s of millions (billions?) of people who spend more than an hour per day just scrolling through social media feeds.
How much does it cost to send a billion people an hour of video every day? Almost all of the resources tech uses is for pointless or even negative things.
What % of compute/bandwidth do you think is used for "real value"? I would guess it is well below 1%.
The pet detection part isn’t the point, that’s just a visible output. The actual goal was to stress test agents + distributed compute on something non-trivial.
Ah yes, let's price the world out of the real estate market and then use insanely powerful AI models to systematically mock the living conditions of the poors.
Fair feedback. Definitely more backlash than I expected. The intent was to experiment with large-scale analysis, not add noise or put strain on shared resources. I’ll be more thoughtful about this kind of thing going forward.
It's content marketing plain and simple for Burla towards people that view this site. It was highly likely done by employees at both Burla and AirBNB together as a joint project.
The branding is a bit much, fair call, but the intent here was just to explore what these agents can actually build when you give them access to large amounts of compute.
This post is a bit lighter on that disclosure than I'd like (and isn't as obvious as a Show HN would be) but I feel I missing some big portion of the backstory to this comment?
As an experitisement, I guess it gets the name out there but not in any way I'd want for my business.
Most of the bad TV placement ones are also boring because they're just over a fireplace. Technically correct, but not noteworthy. However, I did find one that was truly spectacular [1] (still live for now) and left me with more questions than answers.
[0]: https://www.airbnb.com/rooms/988178752120341661 / https://archive.is/xnvC5
[1]: https://www.airbnb.com/rooms/41725492 / https://archive.is/IyMvT
I've had shitty and small apartments many times and that doesn't prevent me from cleaning it. especially if I'm going to rent it out
This Inside Airbnb?
Community Guidelines
Please:
Only take the data you need. Do not scrape data from the site, if you would like to subscribe to the data directly, please email data@insideairbnb.com
That's a lot of budget - would have been nice if they'd made an actual donation to the project, instead of pounding the project's servers and bandwidth when there are much better ways to interact with the data.
I’ll make a donation to support the project regardless. Appreciate you raising it.
>Scale Python across 1,000 CPUs or GPUs in 1 second. Burla is a high-performance parallel processing library with an extremely fast developer experience. Scale batch processing, vector embeddings, inference, or build pipelines with dynamic hardware.
Edit: Author comment was flagged dead. They work at burla which is a managed cloud service for parallelizing python
“Drug-Den vibes” apparently means the owner is poor or a photo is obscured or badly lit.
As an exercise a sample of 1000 photos would've been enough. As a database, knowing a listing has a cat in the picture or a funny review doesn't offer any real value.
I wonder what the footprint is of such an exercise.
How much does it cost to send a billion people an hour of video every day? Almost all of the resources tech uses is for pointless or even negative things.
What % of compute/bandwidth do you think is used for "real value"? I would guess it is well below 1%.
It's hardly the only thing, but it's part of the problem.