Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Infinite Craft

neal.fun

Post reply on HN
Print view View original

291–300 of 831 posts

Re: Infinite Craft

#292

Post by larschdk »

emarthinsen wrote: post #265

Can't believe I was the first to find "boobs". The 12 year-old combined "tomb raider" and "jiggle". LOL. It's actually a useful word. I can change anything male into female. Lots of other funny combinations: Burger King + Business = McDonalds Disney + Motive = Money Princess + Cleavage = Disney Plenty of political ones: President + Dictator = Trump White House + Flame = Trump Trump + Bribe = Impeachment Trump + Impea…

My first discovery was "Santauron"!? - by combining Santa Claus and Sauron.

Re: Infinite Craft

#293

Post by vintermann »

Night_Thastus wrote: post #119
hinkley wrote: post #93

Earlier quoted context omitted.

Forest + fire = smoke Mud + water = swamp Swamp + plant = Venus Flytrap Okay, okay. Venus Flytrap + Smoke = smoke detector

Tree + Water = river was also pretty baffling. Unfortunate. I can see the appeal of using an LLM for this but the results are pretty mediocre.

As usual with language models, you have to put in the work yourself to have fun with them.

I laughed when "Vindaloo + Bubble" gave me "Burp", and "Burp + No Bacon" gave me "Sad"

Re: Infinite Craft

#296

Post by felurx »

jerbear4328 wrote: post #248

I got some funny ones, like: - Bread + Bread = Toast, + Fire = Burnt Toast, + Fire = Ash - Toast + Disaster = Pop Tart, + Fire = Hot Pocket, + Coffee = Coffee Pocket (First Discovery) - Hot Pocket + Cappuccino = Cappuccino Pocket (First Discovery), + Pop Tart = Cappuccino Pop Tart (First Discovery) - Coffee Pocket + Disaster = Coffee Spill (First Discovery), + Pop Art (sic) = Jackson Pollock - Jackson Pollock + Megad…

I played around quite a lot and got many funny combos too. My memory is poor, but one of my favs was Gay + Vampire = Twilight

Re: Infinite Craft

#297

Post by octuplehomicide »

Rich + Hooker = Marriage

Richest + Prostitute = Bill Gates

Bill Gates + Marriage = Divorce

Queen + Divorce = Behead

Trillionaire + Behead = Decapitalism (lol!)

Narcissist + Money = Politician

Peace + Trump = War

Re: Infinite Craft

#298

Post by vintermann »

sva_ wrote: post #237
hinkley wrote: post #93

Earlier quoted context omitted.

Forest + fire = smoke Mud + water = swamp Swamp + plant = Venus Flytrap Okay, okay. Venus Flytrap + Smoke = smoke detector

My favorite was Lotus Flower + Mud = Buddha

Oh, that's a good one! I'm trying to collect religions and countries.

Re: Infinite Craft

#299

Post by schoen »

There are a couple of things that seem to combine with others in a very general way... "were-", "[the] Impaler" (!), "Rich[est]"...

I was able to get some things like "Were-jesus-pope-hamster-crusader", "Trump-jesus-trump", "Pope Impaler", "Trump Impaler", "Vlad the Werehamster", "Richest were-genghis Khan", and also "Were-were-werewolf"!

I think the willingness of the language model to make up character names like this is a bit of a defect in this context, but maybe it's useful for brainstorming movie sequels. :-)

Re: Infinite Craft

#300

Post by adtac »

IshKebab wrote: post #137
leroman wrote: post #63

Earlier quoted context omitted.

I thought they were using some kind of vector space searches like embedding.. no idea if that's the case

Yeah I'm pretty sure you could do this just with the classic word embeddings (king =queen + man - woman). Maybe it doesn't work as well as with a full LLM.

Addition won't work for things that depend on the order of operations. If salt + water is ocean and water + fire is steam, what's salt + water + fire? Is it salt + steam or ocean + fire?

Associativity and commutivity in vector addition doesn't translate well to semantic meaning. Extrapolating your example, it'd also mean:

    v_king  - v_queen   ~= v_father  - v_mother
    v_king  - v_royalty ~= v_father  - v_parent
    v_king  - v_father  ~= v_royalty - v_parent ~= v_queen - v_mother
I don't see why those should all be true. Intuitively, trying to satisfy O(N^2) semantic pairings with vectors that are optimised for a very specific and different numerical operation (cosine similarity) feels like something that won't work. I'd imagine errors get amplified with 3+ operands.
Post reply on HN