> there are approximately 200k common nouns in English, and then we square that, we get 40 billion combinations. At one second per, that's ~1200 years, but then if we parallelize it on a supercomputer that can do 100,000 per second that would only take 3 days. Given that ChatGPT was trained on all of the Internet and every book written, I'm not sure that still seems infeasible.
How would you generate a picture of Noun + Noun in the first place in order to train the LLM with what it would look like? What's happening during that 1 estimated second?
Use any of the image generation models (eg Nanobanana, Midjourney, or ChatGPT) to generate a picture of a noun on a noun. Simonw's test is to have a Language (text) model generate a Scalar Vector Graphic, which the language model has to do by writing curves and colors, like draw a spline from point 150,100 to 200,300 of type cubic, using width 20, color orange.
In that hypothetical second is freaking fascinating. It's a denoising algorithm, and then a bunch of linear algebra, and out pops a picture of a pelican on a bicycle. Stable diffusion does this quite handily. https://stablediffusionweb.com/image/6520628-pelican-bicycle...
The prompt was "a pelican riding a bicycle"; not prepositions but every verb. Potentially every adverb+verb combination - "a pelican clumsily pushing a bicycle"
One thing I'd recommend is getting your own domain for email (looks like mailbox.org allows custom domains with some plans). You never know what will happen to your email provider in the future, so having the freedom to move your domain to a different provider is valuable.
This. This 100%. I hosted my own email for a good few years and decided to migrate it to icloud using the custom domain. A few mins of copying IMAP over and waiting for the DNS to propagate and it was like nothing happened!
It does and I’ve been hoping to see more discussion around best registrars from a domain security perspective. I looked into CSC (cscglobal.com) since it’s what a lot of big companies use, but it’s crazy expensive ($5K+/yr). Even worse, their contract is wild: no guarantee of registration/renewal, all fees non-refundable, they can hike prices anytime, liability capped at $5K, DNS is “as-is” with only credits for outages, and they can unmask WHOIS privacy at will. Basically you pay enterprise prices while they disclaim almost all responsibility.
The source code is such a fun read (for the comments). I found some source code for GW-BASIC, and here are two of my favorites:
;WE COULD NOT FIT THE NUMBER INTO THE BUFFER DESPITE OUR VALIENT
;EFFORTS WE MUST POP ALL THE CHARACTERS BACK OFF THE STACK AND
;POP OFF THE BEGINNING BUFFER PRINT LOCATION AND INPUT A "%" SIGN THERE
;CONSTANTS FOR THE RANDOM NUMBER GENERATOR FOLLOW
;DO NOT CHANGE THESE WITHOUT CONSULTING KNUTH VOL 2
;CHAPTER 3 FIRST
Fun fact, GW-BASIC was a descendant of the original Altair BASIC. The "Translation created 10-Feb-83" headers on each source file refer to tooling Microsoft had that automatically translated the 8080 assembly to 8086 (it shouldn't be taken as a build date since they were manually modified after that point). Besides GW-BASIC, source code for the 6502 and 6809 rewrites of Microsoft BASIC were available up to this point (see https://www.pagetable.com/?p=774 and https://github.com/davidlinsley/DragonBasic) but I believe this is the first public release of the original 8080 BASIC code.
Eclipse has had an integrated spell-checker, which I believe is on by default for most file types, for like approximately forever. Now maybe everybody turns it off, but I gotta imagine there are some people who like it and keep it on.
Emacs has the ability to do spellcheck inline, both as a run through the buffer (old-school style) and as an as-you-type live feature. That said, I do most of my coding in JetBrains IDEs these days.
For Vim/Neovim users, there is one built in that is pretty good, and once you've added frequent custom words to the dictionary it is great. You can turn it on with `:set spell` or off with `:set nospell`. Add custom words by pressing `zg` on the target word:
I have this in my vimrc file so it's on by default for certain file types:
" Turn on spellcheck for certain filetypes and word completion.
" words can be added to the dict by pressing 'zg' with cursor on word.
autocmd Filetype markdown setlocal spell
autocmd Filetype gitcommit setlocal spell
set complete+=kspell
" Don't highlight in red an underscore (_) in markdown
" https://vi.stackexchange.com/q/18471/17441
autocmd Filetype markdown syn match markdownIgnore "\v\w_\w"
Custom additions to the dictionary will go to a simple text file (one word per line) in `~/.vim/spell/en.utf-8.add` (depending on your settings) where it is easy to edit or backup.
You can also add it directly when using it. Move the cursor to the word and (I forget the command...) can add it as a rare word, good word or bad word.
I am far more confident at spelling any Esperanto word that I have never faced before than I am with many common word in French which is my native language.
We can do better than blaming people for falling in pitfalls of a system full of odd traps.
Of course you are; Esperanto is a manufactured language designed to a certain standard; French, like our unfortunate English, is naturally evolved and has all the variants and inconsistencies that implies.
I use a Mac now, but have a Windows VM running in Parallels. Paint.net is one of the first apps I installed on that VM. It's ridiculously easy and intuitive.
The closest program I've found for Mac is Pixelmator Pro, but it's not free, and the UI/workflow is more complex in my opinion. Paint.NET is still one of my favorite pieces of software on Windows.
I spent SO. MUCH. TIME. in computer stores (yes, they were a thing) as a kid in the 80's, and the Compucolor was a favorite. The keys with different colors! Everything was opinionated and geeky. And hardly any demos existed, so the computers were wide open to experiment with.
reply