Getting Started with Celtic Coins – Crude and Barbarous, or Just Different? (collectingancientcoins.co.uk)

I've been playing with embeddings and wanted to try out what results the embedding layer will produce based on just word-by-word input and addition / subtraction, beyond what many videos / papers mention (like the obvious king-man+woman=queen). So I built something that doesn't just give the first answer, but ranks the matches based on distance / cosine symmetry. I polished it a bit so that others can try it out, too.

For now, I only have nouns (and some proper nouns) in the dataset, and pick the most common interpretation among the homographs. Also, it's case sensitive.

Comments (87)

godelski · 2h ago

  data + plural = number
  data - plural = research
  king - crown = (didn't work... crown gets circled in red)
  king - princess = emperor
  king - queen = kingdom
  queen - king = worker
  king + queen = queen + king = kingdom
  boy + age = (didn't work... boy gets circled in red)
  man - age = woman
  woman - age = newswoman
  woman + age = adult female body (tied with man)
  girl + age = female child
  girl + old = female child

The other suggestions are pretty similar to the results I got in most cases. But I think this helps illustrate the curse of dimensionality (i.e. distances are ill-defined in high dimensional spaces). This is still quite an unsolved problem and seems a pretty critical one to resolve that doesn't get enough attention.

n2d4 · 51m ago

For fun, I pasted these into ChatGPT o4-mini-high and asked it for an opinion:

   data + plural    = datasets
   data - plural    = datum
   king - crown     = ruler
   king - princess  = man
   king - queen     = prince
   queen - king     = woman
   king + queen     = royalty
   boy + age        = man
   man - age        = boy
   woman - age      = girl
   woman + age      = elderly woman
   girl + age       = woman
   girl + old       = grandmother

The results are surprisingly good, I don't think I could've done better as a human. But keep in mind that this doesn't do embedding math like OP! Although it does show how generic LLMs can solve some tasks better than traditional NLP.

The prompt I used:

> Remember those "semantic calculators" with AI embeddings? Like "king - man + woman = queen"? Pretend you're a semantic calculator, and give me the results for the following:

franga2000 · 22m ago

This is an LLM approximating a semantic calculator, based solely on trained-in knowledge of what that is and probably a good amount of sample output, yet somehow beating the results of a "real" semantic calculator. That's crazy!

The more I think about it the less surprised I am, but my initial thoughts were quite simply "now way" - surely an approximation of an NLP model made by another NLP model can't beat the original, but the LLM training process (and data volume) is just so much more powerful I guess...

nbardy · 31m ago

I hate to be pedantic, but the llm is definitely doing embedding math. In fact that’s all it does.

gweinberg · 29m ago

I got a bunch of red stuff also. I imagine the author cached embeddings for some words but not really all that many to save on credits. I gave it mermaid - woman and got merman, but when I tried to give it boar + woman - man or ram + woman - man, it turns out it has never heard of rams or boars.

Affric · 1h ago

Yeah I did similar tests and got similar results.

Curious tool but not what I would call accurate.

thatguysaguy · 28m ago

Can you elaborate on what the unsolved problem you're referring to is?

Jimmc414 · 1m ago

dog - cat = paleolith

paleolith + cat = Paleolithic Age

paleolith + dog = Paleolithic Age

paleolith - cat = neolith

paleolith - dog = hand ax

cat - dog = meow

Wonder if some of the math is off or I am not using this properly

montebicyclelo · 1h ago

> king-man+woman=queen

Is the famous example everyone uses when talking about word vectors, but is it actually just very cherry picked?

I.e. are there a great number of other "meaningful" examples like this, or actually the majority of the time you end up with some kind of vaguely tangentially related word when adding and subtracting word vectors.

(Which seems to be what this tool is helping to illustrate, having briefly played with it, and looked at the other comments here.)

(Btw, not saying wordvecs / embeddings aren't extremely useful, just talking about this simplistic arithmetic)

Retr0id · 38m ago

I think it's slightly uncommon for the vectors to "line up" just right, but here are a few I tried:

actor - man + woman = actress

garden + person = gardener

rat - sewer + tree = squirrel

toe - leg + arm = digit

gregschlom · 1h ago

Also, as I just learned the other day, the result was never equal, just close to "queen" in the vector space.

raddan · 1h ago

> is it actually just very cherry picked?

100%

groby_b · 34m ago

I think it's worth keeping in mind that word2vec was specifically trained on semantic similarity. Most embedding APIs don't really give a lick about the semantic space

And, worse, most latent spaces are decidedly non-linear. And so arithmetic loses a lot of its meaning. (IIRC word2vec mostly avoided nonlinearity except for the loss function). Yes, the distance metric sort-of survives, but addition/multiplication are meaningless.

(This is also the reason choosing your embedding model is a hard-to-reverse technical decision - you can't just transform existing embeddings into a different latent space. A change means "reembed all")

rdlw · 19m ago

I've always wondered if there's s way to find which vectors are most important in a model like this. The gender vector man-woman or woman-man is the one always used in examples, since English has many gendered terms, but I wonder if it's possible to generate these pairs given the data. Maybe to list all differences of pairs of vectors, and see if there are any clusters. I imagine some grammatical features would show up, like the plurality vector people-person, or the past tense vector walked-walk, but maybe there would be some that are surprisingly common but don't seem to map cleanly to an obvious concept.

Or maybe they would all be completely inscrutable and man-woman would be like the 50th strongest result.

__MatrixMan__ · 34m ago

Here's a challenge: find something to subtract from "hammer" which does not result in a word that has "gun" as a substring. I've been unsuccessful so far.

mrastro · 27m ago

The word "gun" itself seems to work. Package this as a game and you've got a pretty fun game on your hands :)

neom · 27m ago

if I'm allowed only 1 something, I can't find anything either, if I'm allowed a few somethings, "hammer - wine - beer - red - child" will get you there. Guessing given that a gun has a hammer and is also a tool, it's too heavily linked in the small dataset.

Retr0id · 27m ago

Well that's easy, subtract "gun" :P

downboots · 25m ago

Bullet

tough · 27m ago

hammer + man = adult male body (75%)

spindump8930 · 2h ago

First off, this interface is very nice and a pleasure to use, congrats!

Are you using word2vec for these, or embeddings from another model?

I also wanted to add some flavor since it looks like many folks in this thread haven't seen something like this - it's been known since 2013 that we can do this (but it's great to remind folks especially with all the "modern" interest in NLP).

It's also known (in some circles!) that a lot of these vector arithmetic things need some tricks to really shine. For example, excluding the words already present in the query[1]. Others in this thread seem surprised at some of the biases present - there's also a long history of work on that [2,3].

[1] https://blog.esciencecenter.nl/king-man-woman-king-9a7fd2935...

[2] https://arxiv.org/abs/1905.09866

[3] https://arxiv.org/abs/1903.03862

nxa · 2h ago

Thank you! I actually had a hard time finding prior work on this, so I appreciate the references.

The dictionary is based on https://wordnet.princeton.edu/, no word2vec. It's just a plain lookup among precomputed embeddings (with mxbai-embed-large). And yes, I'm excluding words that are present in the query because.

It would be interesting to see how other models perform. I tried one (forgot the name) that was focused on coding, and it didn't perform nearly as well (in terms of human joy from the results).

kaycebasques · 1h ago

(Question for anyone) how could I go about replicating this with Gemini Embedding? Generate and store an embedding for every word in the dictionary?

nxa · 1h ago

Yes, that's pretty much what it is. Watch out for homographs.

antidnan · 3h ago

Neat! Reminds me of infinite craft

https://neal.fun/infinite-craft/

thaumasiotes · 43m ago

I went to look at infinite craft.

It provides a panel filled with slowly moving dots. Right of the panel, there are objects labeled "water", "fire", "wind", and "earth" that you can instantiate on the panel and drag around. As you drag them, the background dots, if nearby, will grow lines connecting to them. These lines are not persistent.

And that's it. Nothing ever happens, there are no interactions except for the lines that appear while you're holding the mouse down, and while there is notionally a help window listing the controls, the only controls are "select item", "delete item", and "duplicate item". There is also an "about" panel, which contains no information.

n2d4 · 41m ago

In the panel, you can drag one of the items (eg. Water) onto another one (eg. Earth), and it will create a new word (eg. Plant). It uses AI, so it goes very deep

thaumasiotes · 39m ago

No, that was the first thing I tried. The only thing that happens is that the two objects will now share their location. There are no interactions.

n2d4 · 36m ago

Probably a bug then, you can check YouTube to find videos of people playing it (eg. [0])

[0] https://youtu.be/8-ytx84lUK8

lcnPylGDnU4H9OF · 1h ago

Some of these make more sense than others (and bookshop is hilarious even if it's only the best answer by a small margin; no shade to bookshop owners).

  map - legend = Mercator projection
  noodle - wheat = egg noodle
  noodle - gluten = tagliatelle
  architecture - calculus = architectural style
  answer - question = comment
  shop - income = bookshop
  curry - curry powder = cuisine
  rice - grain = chicken and rice
  rice + chicken = poultry
  milk + cereal = grain
  blue - yellow = Fiji
  blue - Fiji = orange
  blue - Arkansas + Bahamas + Florida - Pluto = Grenada

jumploops · 1h ago

This is super neat.

I built a game[0] along similar lines, inspired by infinite craft[1].

The idea is that you combine (or subtract) “elements” until you find the goal element.

I’ve had a lot of fun with it, but it often hits the same generated element. Maybe I should update it to use the second (third, etc.) choice, similar to your tool.

[0] https://alchemy.magicloops.app/

[1] https://neal.fun/infinite-craft/

nxa · 1h ago

This might be helpful: I haven't implemented it in the UI, but from the API response you can see what the word definitions are, both for the input and the output. If the output has homographs, likeliness is split per definition, but the UI only shows the best one.

Also, if it gets buried in comments, proper nouns need to be capitalized (Paris-France+Germany).

I am planning on patching up the UI based on your feedback.

neom · 34m ago

cool but not enough data to be useful yet I guess. Most of mine either didn't have the words or were a few % off the answer, vehicle - road + ocean gave me hydrosphere, but the other options below were boat, ship, etc. Klimt almost made it from Mozart - music + painting. doctor - hospital + school = teacher, nailed it.

Getting to cornbread elegantly has been challenging.

lightyrs · 2h ago

I don't get it but I'm not sure I'm supposed to.

    life + death = mortality
    life - death = lifestyle

    drug + time = occasion
    drug - time = narcotic

    art + artist + money = creativity
    art + artist - money = muse

    happiness + politics = contentment
    happiness + art      = gladness
    happiness + money    = joy
    happiness + love     = joy

bee_rider · 1h ago

    Life + death = mortality

is pretty good IMO, it is a nice blend of the concepts in an intuitive manner. I don’t really get

   drug + time = occasion

But

   drug - time = narcotic

Is kind of interesting; one definition of narcotic is

> a drug (such as opium or morphine) that in moderate doses dulls the senses, relieves pain, and induces profound sleep but in excessive doses causes stupor, coma, or convulsions

https://www.merriam-webster.com/dictionary/narcotic

So we can see some element of losing time in that type of drug. I guess? Maybe I’m anthropomorphizing a bit.

grey-area · 2h ago

Does the system you’re querying ‘get it’? From the answers it doesn’t seem to understand these words or their relations. Once in a while it’ll hit on something that seems to make sense.

yigitkonur35 · 22m ago

shows how bad embeddings are in a practical way

grey-area · 2h ago

As you might expect from a system with knowledge of word relations but without understanding or a model of the world, this generates gibberish which occasionally sounds interesting.

afandian · 1h ago

There was a site like this a few years ago (before all the LLM stuff kicked off) that had this and other NLP functionality. Styling was grey and basic. That’s all I remember.

I’ve been unable to find it since. Does anyone know which site I’m thinking of?

halter73 · 1h ago

I'm not sure this is old enough, but could you be referencing https://neal.fun/infinite-craft/ from https://news.ycombinator.com/item?id=39205020?

ericdiao · 2h ago

Interesting: parent + male = female (83%)

Can not personally find the connection here, was expecting father or something.

ericdiao · 2h ago

Though dad is in the list with lower confidence (77%).

High dimension vector is always hard to explain. This is an example.

matallo · 1h ago

uncle + aunt = great-uncle (91%)

great idea, but I find the results unamusing

HWR_14 · 59m ago

Your aunt's uncle is your great-uncle. It's more correct than your intuition.

matallo · 42m ago

I asked ChatGPT (after posting my comment) and this is the response. "Uncle + Aunt = Great-Uncle is incorrect. A great-uncle is the brother of your grandparent."

skeptrune · 2h ago

This is super fun. Offering the ranked matches makes it significantly more engaging than just showing the final result.

downboots · 29m ago

three + two = four (90%)

LadyCailin · 26m ago

Haha, yes, this was my first thought too. It seems it’s quite bad at actual math!

cabalamat · 2h ago

What does it mean when it surrounds a word in red? Is this signalling an error?

iambateman · 2h ago

Try Lower casing, my phone tried to capitalize and it was a problem.

fallinghawks · 2h ago

Seems to be a word not in its dictionary. Seems to not have any country or language names.

Edit: these must be capitalized to be recognized.

nxa · 2h ago

Yes, word in red = word not found mostly the case when you try plurals or non-nouns (for now)

rpastuszak · 2h ago

This is neat!

I think you need to disable auto-capitalisation because on mobile the first word becomes uppercase and triggers a validation error.

tlhunter · 44m ago

man + woman = adult female body

kylecazar · 1h ago

Woman + president = man

fallinghawks · 2h ago

goshawk-cocaine = gyrfalcon , which is funny if you know anything about goshawks and gyrfalcons

(Goshawks are very intense, gyrs tend to be leisurely in flight.)

ericdiao · 1h ago

wine - alcohol = grape juice (32%)

Accurate.

nikolay · 2h ago

Really?!

  man - brain = woman
  woman - brain = businesswoman

nxa · 2h ago

I probably should have prefaced this with "try at your own risk, results don't reflect the author's opinions"

dmonitor · 37m ago

I'm sure it would be trivial to get it to say something incredibly racist, so that's probably a worthwhile disclaimer to put on the website

dalmo3 · 2h ago

I think subtraction is broken. None of what I tried made any sense. Water - oxygen = gin and tonic.

sapphicsnail · 2h ago

Telling that Jewess, feminist, and spinster were near matches as well.

karel-3d · 2h ago

woman+penis=newswoman (businesswoman is second)

man+vagina=woman (ok that is boring)

2muchcoffeeman · 2h ago

Man - brain = Irish sea

nikolay · 2h ago

Case matters, obviously! Try "man" with a lower-case "M"!

Alifatisk · 2h ago

Why does case matter? How does it affect the meaning?

bfLives · 2h ago

“Man” is probably being interpreted as the Isle of Man.

https://en.m.wikipedia.org/wiki/Isle_of_Man

G1N · 2h ago

Man (capital M) is probably being interpreted as some proper noun, maybe Isle of Man in this case?

MYEUHD · 2h ago

king - man + woman = queen

queen - woman + man = drone

bee_rider · 1h ago

The second makes sense, I think, if you are a bee.

firejake308 · 3h ago

King-man+woman=Navratilova, who is apparently a Czech tennis player. Apparently, it's very case-sensitive. Cool idea!

fph · 2h ago

"King" (capital) probably was interpreted as https://en.wikipedia.org/wiki/Billie_Jean_King , that's why a tennis player showed up.

nxa · 2h ago

when I first tried it, king was referring to the instrument and I was getting a result king-man+woman=flute ... :-D

BeetleB · 2h ago

Heh. This is fun:

Navratilova - woman + man = Lendl

zerof1l · 2h ago

male + age = female

female + age = male

blobbers · 2h ago

rice + fish = fish meat

rice + fish + raw = meat

hahaha... I JUST WANT SUSHI!

TZubiri · 1h ago

I'm getting Navralitova instead of queen. And can't get other words to work, I get red circles or no answer at all.

gus_massa · 1h ago

From another comment, https://news.ycombinator.com/item?id=43988861 King (with capital K) was a top 1 male tenis player.

7373737373 · 2h ago

it doesn't know the word human

adzm · 2h ago

noodle+tomato=pasta

this is pretty fun

growlNark · 2h ago

Surely the correct answer would be `pasta-in-tomato-sauce`? Pasta exists outside of tomato sauce.

kataqatsi · 2h ago

garden + sin = gardening

hmm...

G1N · 2h ago

twelve-ten+five=

six (84%)

Close enough I suppose

woodruffw · 2h ago

colorless+green+ideas doesn't produce anything of interest, which is disappointing.

dmonitor · 36m ago

well green is not a creative color, so that's to be expected

ezbie · 1h ago

Can someone explain me what the fuck this is supposed to be!?

mhitza · 41m ago

Semantical subtraction within embeddings representation of text ("meaning")

LLMs Are Making Me Dumber (vvvincent.me)

AlphaEvolve: A Gemini-powered coding agent for designing advanced algorithms (deepmind.google)

Show HN: Muscle-Mem, a behavior cache for AI agents (github.com)

What is HDR, anyway? (lux.camera)

Migrating to Postgres (engineering.usemotion.com)

Show HN: Semantic Calculator (king-man+woman=?) (calc.datova.ai)

A server that wasn't meant to exist (it-notes.dragas.net)

Git Bug: Distributed, Offline-First Bug Tracker Embedded in Git, with Bridges (github.com)

Getting Started with Celtic Coins – Crude and Barbarous, or Just Different? (collectingancientcoins.co.uk)

Smalltalk-78 Xerox NoteTaker in-browser emulator (smalltalkzoo.thechm.org)

StackAI (YC W23) Is Hiring Pydantic and FastAPI Wizard (ycombinator.com)

Variadic Switch (pydong.org)

Our narrative prison (aeon.co)

Hegel 2.0: The imaginary history of ternary computing (2018) (cabinetmagazine.org)

The cryptography behind passkeys (blog.trailofbits.com)

Changes since congestion pricing started in New York (nytimes.com)

Databricks and Neon (databricks.com)

The AUCTUS A6: the chip enabling inexpensive DMR Radio (2021) (jhart99.com)

Bus stops here: Shanghai lets riders design their own routes (sixthtone.com)

Perverse incentives of vibe coding (fredbenenson.medium.com)

UK's Ancient Tree Inventory (ati.woodlandtrust.org.uk)

The recently lost file upload feature in the Nextcloud app for Android (nextcloud.com)

How the economics of multitenancy work (blacksmith.sh)

Updated rate limits for unauthenticated requests (github.blog)

An accessibility update – GTK Development Blog (blog.gtk.org)

Launch HN: Jazzberry (YC X25) – AI agent for finding bugs

Interferometer Device Sees Text from a Mile Away (physics.aps.org)

Show HN: Lumier – Run macOS VMs in a Docker (github.com)

How to Build a Smartwatch: Picking a Chip (ericmigi.com)

We Made CUDA Optimization Suck Less (rightnowai.co)

Various Things in MetaPost (2019) (habr.com)

Uber to introduce fixed-route shuttles in major US cities (techcrunch.com)

Writing that changed how I think about programming languages (bernsteinbear.com)

Show HN: CSV GB+ by Data.olllo – Open and Process CSVs Locally (apps.microsoft.com)

Ash Framework – Model your domain, derive the rest (ash-hq.org)

SMS 2FA is not just insecure, it's also hostile to mountain people (blog.stillgreenmoss.net)

The pigeon whistle: A defining sound of old Beijing (2019) (chinatoday.com.cn)

Flattening Rust’s learning curve (corrode.dev)

Show HN: I’ve built an IoT device to let my family know when I’m in a meeting (nullonerror.org)

Finding Beauty and Truth in Mundane Occurrences (quantamagazine.org)

The overlooked masterpiece full of coded messages about World War One (bbc.com)

RPG in a Box (rpginabox.com)

Type-constrained code generation with language models (arxiv.org)

Breaking the Sound Barrier Part I: Fuzzing CoreAudio with Mach Messages (googleprojectzero.blogspot.com)

David Hilbert's radio address (2014) (old.maa.org)

TwoFold (2f), CLI text expander/template engine (github.com)

Launch HN: Miyagi (YC W25) turns YouTube videos into online, interactive courses

E-COM: The $40M USPS project to send email on paper (buttondown.com)

Branch Privilege Injection: Exploiting branch predictor race conditions (comsec.ethz.ch)

AI Agents Must Follow the Law (lawfaremedia.org)

Show HN: Semantic Calculator (king-man+woman=?)

Comments (87)