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

Author here! Funny to see this at the top of HN today -- happy to answer any questions (source code is here https://github.com/turtlesoupy/this-word-does-not-exist)

Shameless plug for my other "this x does not exist": This Fucked Up Homer Does Not Exist https://www.thisfuckeduphomerdoesnotexist.com/

Ahh, and you've made the wise decision to NFTify the fucked up Homers! Nothing says "investment vehicle" like-- hey, wait a minute!! That's a fucked up Bart!

Good day, sir!

[0] https://opensea.io/assets/0x495f947276749ce646f68ac8c2484200...

Alright! Someone shilling on my behalf. Say what you want about NFTs, every time I see one of the homers sell it brings a big smile to my face.

Frankly, I'm mocking it. Or trying to. Riffing on it? NFTs strike me as obvious nonsense -- why not use all that energy for something more useful, like piping /dev/urandom to /dev/null -- but NFT sales of obvious nonsense, of an algorithmic corruption of an artistic representation of a shared cultural totem, are almost high art.

Like, unpleasantly high art. Way too high art. Art that makes me want to go lie down for a bit.

It's just digital trading cards, like pokemon or baseball cards. Just amateurs are running the show now, but the NBA already got in on this, so if you want something more legit: https://nbatopshot.com/

Hey now, they're not just "digital trading cards, like pokemon or baseball cards"! They're also an enormous waste of resources!

Yeesh. At least collectors of sports memorabilia have some memorabilia to admire. I can't imagine what value-add there is for "NBA Top Shots" that I can't replicate by backing up a short clip of the play. My solution is more durable, too!

All of money is imaginary and all things have a representative monetary value. All of the above are imaginary things. Work from this basis, the rest of this will make more sense. So now if you use your imagination, we can start lining up some value. What value does gold have in World of Warcraft?

You can admire your digital trading cards! Just check out this program I made that cycles your NFT's as background wallpapers for your computer. github.com/NotaRealThing /s

Gold is just a rock. Why do all humans value that nonsense so much?

No, gold is an amazing substance in its own right. There are rarer and scarcer metals, but gold is special. This conversation relies in some part on the ductility and conductivity of gold to power our computers. It doesn’t tarnish and it is very malleable. This means you can shape it into some very beautiful and durable forms with even crude hand tools. The potential value add is very large. It’s use to back currencies is largely symbolic but it was chosen for good reasons.

Gold is also impractical to synthesize by modern methods!

Also it is very dense, silver is not so it could be forged too easy.

Gold has properties that are similar to copper (highly electrically and thermally conductive, lustrous, and malleable), with the added benefit of being corrosion resistant. Many of the industrial applications of gold and copper are similar.

Which would suggest that if gold and copper had the same supply, the demand would be similar and thus the price would be the same. It turns out that's approximately correct— the ratio of gold and copper's prices is similar to the ratio of their supply.

> waste of resources

Are you able to provide data comparing the footprint of NFT's to baseball cards, Magic the Gathering cards, paintings, etc?

Do you even know what you're talking about? Or just regurgitating some talking points you heard somewhere?

No, it's digital money laundering

Lately HN feels like just recycled Matt Levine.

Which is high praise (except for the recycled part).

Data art is anti-dada.

> Frankly, I'm mocking it. Or trying to. Riffing on it?

Frankly, you failed :-)

I'm as sceptic to crypto as almost anyone but this was one of the nicer ideas.

No I'm pretty sure he did fine, NFTs are little more than a vehicle for money laundering and a joke for a normal people.

These are awesome! Thanks for putting your work out into the world, turtlesoup.

In my free time I make paintings of outputs from GANs/Deepdreams/etc. I've been wanting to paint something based on a model trained on illustrations, but I hadn't seen anyone decent GAN trained on that kind of dataset yet. Thanks for sharing your code, this is really exciting to see!

I haven't either (beyond anime / animals, which is widely explored ala https://www.thiswaifudoesnotexist.net/) -- if you do see anything let me know.

Tangentially related, one more of my favourite parts of this work an artist that made illustrations of some of the fucked homers: https://twitter.com/turtlesoupy/status/1349541080719478784

Given the rise of Loot and other text-based NFTs, you might want to NFTify this too. I know Loot does this by creating a URL with the text in it that points to a text image generator.

Which is a bad strategy as dictated by signaling theory [1]. There's nothing rare or costly about them when you can keep hitting refresh to get a new one.

Instead, if there were only a few produced in the world, and you owning it give you some kind of status, then you'd have a real potential NFT.

[1] https://en.wikipedia.org/wiki/Signalling_theory

It is worse. Anyone can own it. NFTs don't avoid copying, and they are in no way recognized as copyrighted material. Its like having a proof of ownership of a public good: worthless.

I don't know what you're talking about.

I just bought the NFT for the SF Golden Gate bridge, which surely means I own it.

The fellow who sold it to me said so.

I bet with the right license you could convey legal rights to the NFT owner.

You might need some filtering. I got the following output for my 7th or 8th try (I did s/ig/*/g because I hate typing this):




a bar from which the beer and liquor are sold

"that party got a whole lot cooler this week with a n*gerbar refreshment"

a word that does not exist; it was invented, defined and used by a machine learning algorithm.

Ah crap, good call. I did a first-pass at filtering for the ones displayed on the site but my regex must have had some misses.

Not as bad, but also probably something for a "blocklist": cockshop a shop selling flannel and scarves

It's hilarious that not even this GPT side project can escape the need for OpenAI-style output filtering.

Someone really ought to make an open source lib to solve this once and for all...

filtering objectionable content actually requires you to build a strong AI model capable of being offended itself, so it knows to hold its tongue in mixed company

edit: lest i leave this comment totally useless, the chatbot engine “chatscript” has pretty good capabilities for disambiguating word meaning and classifying the meanings into “badword” and “verybadword” - its free/libre software and very high performance.


Objectionable is subjective.

So is skillful language use. We nonetheless think it is achievable without using wetware.

Or, better yet: we should all work to make sure every human being has their basic needs met and is treated with respect, and then words like this wouldn't have as much power.

I think actually solving this problem is somewhat loosely reducible to human level intelligence in language understanding. The next best thing is a pile of patches that fix cases of the increasing creativity of the human adversaries to the system as we become aware of them.

You should aim higher than that. After all it was human level intelligence that got a proffesor fired for saying "nèige", a Mandarin filler word, in a lecture about filler words in other languages!


I don’t have access to gpt-3 but have wondered how much adding “without being offensive” to the prompt would help.

On one hand it’s the promise of the algorithm that such things will work, on the other hand I doubt it.

a simple blocklist might be just the cheaper and easier solution. After all, blocklists were, and still are being used to filter human output as well. And fail-safe. A person with a hobby project does not have a legal/pr department to deal with the consequences of AI having a bad day

(you might want to read up on the use-mention distinction https://en.wikipedia.org/wiki/Use%E2%80%93mention_distinctio...)

I have a feeling the parent poster realized they were only mentioning the word, but chose not to write it anyway. For a lot of people, it is a very uncomfortable word after all.

Please could you explain the relevance of the article to the previous comment?

Perhaps throw10920 thinks I should have just used the word because I was only quoting the output, so it wasn't really me saying it, but I chose otherwise. I don't need to read up on the use-mention distinction to decide whether to make such a choice.

Has it generated a word for the noise you make after sipping a cup of tea or a cold drink? The “ahhhh” sound. I nominate the word “fonce” if it’s not already taken.

“The pair sat foncing so noisily over their hot cups of tea that it drove everyone else from the room”

moaning, the word is moaning, or involuntary escaping of air. If it's done really loud and on ourpose that's just called being an asshole, basically.

Ahahahahahaha! What a foncy sippers!

You may have just done more to bastardize the english language than all the slang of the past 10 years.

And by that I mean that the words are good enough that people will use them and lexicographers will have no choice but to add them to dictionaries.

Regrettably I hit refresh, but one word it said did not exist was "geochronology"


It writes Wikipedia articles too? Thus the Singularity begins.

It's got a good grasp on the difference between Norwegian and Swedish, I see; it defines Norwegian as "very light, quiet, or peaceful; tranquil", as in "we live our lives Norwegian in a quiet quiet city", and Swedish as "wicked or excessively wicked", as in "his swedish exploits". Haha...

> https://www.thisfuckeduphomerdoesnotexist.com

I'm now more interested in how the transitions are made than in the generators per se.

I'd suggest this code to follow: https://github.com/lucidrains/stylegan2-pytorch/blob/fc22408.... The standard practice is to interpolate between the source latents and target latents and feed the interpolated values into your GAN. There are various ways of doing interpolation (i.e. not always straight linear interpolation)

No specific question, but thank you for the highly entertaining (and interesting) site; I've had a blast with it.

"A Way With Words" would probably love this.

There's a link called "link" which creates a permalink to a given word.

The alternate site is particularly relevant to me; the first two words I "made up" were "cromulent" and "embiggen".

Well played.

Should have named it startupnamegenerator.com

I think what it does with real words is almost more interesting. The definition it picked for the word "real" was accurate, but kind of odd. If I recall correctly, "real" was defined as, "committed", or "dedicated", as in "real fans only like the vinyl edition of that record." Which is like, true for sure, but not in the top 10 things I'd think of first.

I agree. It does the same with for instance "ballsy", which it defines as "stunning and stylish", as in "the ballsy pink dress".

I just thought of it. Kind of terrific crossword clues.

The "word" segmentation seems pretty off, even for these hallucinated words and meanings. Are they generated by the same model?

What's the meaning of "liff"?

thiscrossworddoesnotexist could be fun and perhaps devilish.

So I just had a hit on a real word (pejorative), not sure if that's the plan or not


The word linked above is "perjorative", which is one letter away from being a real word.

It would be glorious if we could share definitions. Who wouldn't want to tweet "chadocracy"? or "prospectivism, the belief that an action will have a good outcome".

Always curious of how you walk around the latent space in these, are you just linearly interpolating between two random high dimension points?

How long did it take to train this GPT-2 based model?

It's a refinement of a lightweight version of GPT-2 by Hugging Face -- https://huggingface.co/transformers/model_doc/gpt2.html. I don't recall exact numbers, but once I had the structure of the problem right (i.e. sequencing words, part of speech and definitions) it was around 12 hours on my old 1080 TI.

I first got 'rudimentally' which certainly does already exist although it's not the most popular of words.

This is cracking me up LOL

It would be an amazing screensaver, just iterations upon iterations of Fucked Up Homers.

Fantastic work! How did you learn to make the Homers? I would love to create something like this.

Do you do inference on every visit?

What is the cost of being on the front page?

hmm. out of 5 words i generated, two actually exist: bioamine and bioprocessor.

Please say you take requests.

Press "write your own" and it'll make up a definition for you :)


Somewhat inspired by your homer work :)


does the definition exist or is that machine-generated too?

Well duh obviously it exists, arent you reading it? Therefore it falls in the realm of existence

lol jk, im just fucking around with the semantics of words here. However I have to admit, this thought about existence or truthfulness of concepts or statements being spitted out by unsentient machines that mix and match infinite "real" patterns, has me quite worried...

Sometimes I find myself reading at a whole discussion thread, and I get the uncomfortable sensation that everything I've been reading is a bunch of bots training each other's models...

Not quite when the concepts are complex enough, all right, it is easy to spot pointless mouthfuls deviations of a main subject being discussed, but what about, for example, the user reviews on an amazon product? Or a youtube's vid comment section? Or whatever shit you consume from the internet.

I'm by no means any expert in the subject, and current state of the art of the Turing test, but I've seen just enough GPT-3 whichcraft to start being totally skeptical about anything I see online.

sorry for the huge rant, but i took a great effort to make it sound like its coming from a real person XD

also, to @turtlesoup: Thank you, that was wonderful. I hope you enjoy this cup of coffee too :-)

Both definitions and words are machine generated!


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