Show HN: Build a website in 30 seconds using GPT-3 | Hacker News Reader

Show HN: Build a website in 30 seconds using GPT-3

durable.co

49 pointsrookhack20 comments

Just tried it out I am quite impressed with the demo. I tried a non-given buisness prompt "massage parlor" and it did generate results. Although the provided pictures were a bit strange and the initial description of the business said it closed in 2019. But at the footer of the page it said it started in 2020.

I think this is definitely a step up from "lorem ipsum" default text. The generated text was rather formulaic though. For the "services" section every paragraph started with "is a type of massage that".

Thanks for giving it a try! Definitely a lot of polishing to do.

Sometimes the results get super weird (massage parlor is a bit of a tough one), but you can regenerate as many times as you like.

I gave it something I thought would be a "tougher" one than a massage parlor: a flying car dealership, and it handled it great outside of obviously not having access to a real flying car picture. The "about us" generated more of a testamonial, but it was coherent and on topic.

Very awesome, this is great for small businesses and such that don't want to take the time to make something simple and essentially is read only.

Really good job so far. I can think of all kinds of ideas for this.

I have a lot of questions, particularly about how and if you can guide the models to keep pace with design advancements, or is there worry that this would just reinforce the models to make the same design? Maybe you could just have an update button that would users could press every few years to keep up appearances.

I'd also love if these models could be trained to help weed out bad development practices, particularly around the attention economy.

It also makes me think.. like what happens when everything we see online is generated by AI? Every website, comment, video, image, everything and everyone will be in their own little silo.

Are you worried at all about being an engine for outputting large amounts of low quality content?

I can see the appeal of having a fast cookie-cutter design, and generated prompts to instruct the user on the type of content they should include. But looking at the demo, you generate fake testimonials, and loads of generic information-less paragraphs. This raises some ethical questions if you let that go out as is.

Do you have features to incite the users to overwrite those?

Am quietly impressed.

My record label helped put punk on the map in the southwest of England! Shame it closed in 2012, according to the blurb. I really thought we were going places.

How well does this website work with open source GPT-3 alternatives? Did you give those a try?

We haven't tried any yet, but would love to. Stable Diffusion for images would be super cool. Which would you recommend?

I was curious about how GPT-NeoX would match up to GPT-3. It seems like the most capable of the freely available models (at least based on their specs). Of course more parameters != quality of their dataset, so I can't really conclude much in this scenario.

As far as I know, https://textsynth.com is the cheapest host and NeoX-20B model is cheaper than OpenAI's Curie. They have a Playground which allows you to experiment without an account.

Re Stable Diffusion:

Based on the results I've seen in https://news.ycombinator.com/item?id=33038117 and other services, if you just want some eye candy I think Stable Diffusion would work fine (and is much better looking than anything DALL-E could generate).

However I wasn't impressed by SD's ability to understand the meaning of my sentence, it seems to just create an image by mashing words together (and sometimes even ignores key words/meanings). According to commenters, DALL-E is much better at "understanding" sentences so if that matters for this site; then I guess you should avoid SD.

awesome - appreciate that! Will check out textsynth.

We would LOVE a way to generate logos effectively, maybe Stable Diffusion is the answer there. Right now DALL-E really struggles with that.

Logos are really hard though.

Check out the free, open source https://stablehorde.net/ for stable diffusion image generation via a distributed cluster (with REST API).

I disagree with this, I don't think it's a good look to use a free service with limited resources for something commercial; it's just bound to cause conflicts.

But this service (and their sister project for text generation) is incredibly cool and interesting, surprised I haven't heard about it sooner.

Impressed with the results I got. You only allow US citizens to sign up?

Right now for the full product (Durable.co) it includes a business checking account which is limited to the US - but everything else is available worldwide.

You can still sign up and use it, we don't technically limit anything geographically yet (except for that one feature).

Would this cost a lot to run the servers (I’ve heard that the AI is resouce intensive). Then is it worth the money?

It's surprisingly inexpensive! Our bill for the last three days (about 8k websites built) was around $75. Definitely more expensive than standard servers, but not too bad.

Debuild.co basically joined YC on this premise when GPT-3 launched. Not sure how far along they are though.

i love all this "AI" shite and cliche web design because it has a distinct visual signature that helps me quickly identify low quality garbage.

thanks for checking it out. Our users aren't discerning design experts like yourself, but mostly small owner-operator businesses. The websites they make here are usually way better than their current one, that they've been meaning to update for years but don't have the time to.

Pretty cool !

Reply on news.ycombinator.com