Hacker News new | past | comments | ask | show | jobs | submit login

There's still plenty of incredible hand-made and vintage furniture and interior design objects on Etsy.

The problem, as ever, is about filtering. For every quality piece there 50 amateurish things cobbled together from steel pipes and reclaimed wood (or worse: distressed "shabby chic" furniture made to look like something you'd find in an old B&B on the French Riviera). With all the AI being thrown around, can't someone build a "tastefulness" filter?

You're right about great furniture and objects. We've tried putting a local focus on some of their selection (as well as the selection from additional non-etsy sellers), by selecting/curating furniture dealers from the DC and Baltimore areas so that would-be buyers can have a look in person:


Especially in regards to furniture, buying online, sight unseen still isn't a great proposition. We don't expect it will ever be.

Perhaps with AR or VR but it's still hard to imagine supplanting the experience without an actual tactile connection. Furniture isn't something you buy everyday or even year typically. If I am going to spend any amount of money on a couch or chair or whatever I'd like to sit in it atleast once before I do (though perhaps that's just me others may be different).

Perhaps though online could be a great first step filtering process. Find a bunch of stuff that fits the rough parameters of what you are looking for and looks good in your space then go someplace to try them out with a well kept appointment and ready to go same day follow you home delivery should you decide to buy something.

Maybe this already exists? I haven't had to buy any furniture is several years so not sure what the current state of the markets is.

Cool. Would love something like this for NYC.

> With all the AI being thrown around, can't someone build a "tastefulness" filter

I've long wanted an extension for Google Image Search, eBay, Etsy, etc. that simply filtered out all listings with images of items with pure white backgrounds. That would go a long ways toward a "tastefulness filter".

Huge indicator of handmade / vintage / used items sold by an individual seller vs. a new item being dropshipped from Amazon or China.

Sadly, they sort of figured that out already: http://www.storenvy.com/shop/category/home-furniture/. (Yes, that's their furniture category. Who knew ugly, amateurishly made dolls were considered furniture?)

Are the hands fake too?!

I like the idea, but this would also be easy to game.

These guys are using deep learning/convolutional neural networks to train their models to detect beautiful/thoughtful photographic images.


They talk about their curation algorithms here:


> With all the AI being thrown around, can't someone build a "tastefulness" filter?

It turns out taste varies.

I would argue (somewhat tongue in cheek) that people who don't have a taste [1] don't need a filter. The filter is for people who want to climb above the trash, not for those who willingly submerge themselves.

[1] https://www.etsy.com/shop/GrainyDayCreations, https://www.etsy.com/shop/PurpleHeartUK, https://www.etsy.com/shop/BigSwigDesign, https://www.etsy.com/shop/CraftyMamaGifts, etc.

Right. I wouldn't use that; while I can totally understand why you linked the shops you did, they also aren't qualitatively much different than the shops I do appreciate. So I'd rather not automate away my taste until, you know, it's my taste.

Based on those links, I think you may have confused "taste" with "irony". :)

Still sounds like an interesting problem.

Perhaps a set of AI filters could be built, and Netlix or Pandora style whichever one a person agrees with more often is selected and more results from it are presented to them.

Or perhaps a Convolution Network trained to filter per person or something equally wasteful if the allocations of Etsy's budget actually match the article. But of course those GPUs would be locally sourced and powered by renewable hamster wheels and 100% organic hamsters.

Which just means that taste is a function of a person. Or, approximately, a function of all the data you have on them.

Registration is open for Startup School 2019. Classes start July 22nd.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact