You Autocomplete Me: Stable Diffusion—What’s The Big Deal?
--
For anyone who remembers typing TURN OFF LIGHT SWITCH while playing the text-adventure game Zork—instead of typing 10 PRINT "HELLO WORLD" because there were no programs to run besides the ones you could make yourself—you recall the feeling of being drawn into an alternate universe.
I think it’s because back then we were so accustomed to talking to the computer in its language—the language of the machine—that it felt odd and wonderful to speak to it in our own language. In my recent book, How to Speak Machine, I provide a gentle introduction to AI/ML for anyone who doesn’t natively “speak machine.” Because you’re more likely to speak in regular human language. It’s what makes reading what I’m typing so much easier to understand than some math-y gobbledygook.
When we think, it has a certain feeling. And that feeling is less likely to look like computer code … say for counting to a million:
void main()
{
int i;
for(i = 0; i < 1000000; i++)
printf("%d",i);
}That’s how a machine likes to think: in terms of loops. Preferably ones with infinite bounds so that it can repeat itself precisely over and over and over and over. In contrast, we humans think less robotically and more associatively.
You Autocomplete Me**
The best depiction I’ve ever found for how that feels is in BBC’s newer rendition of Sherlock Holmes starring Benedict Cumberbatch. There are particular scenes in which the main protagonist eventually arrives to his “Elementary!” moment where Holmes struggles through various concepts that he’s attempting to formulate—almost like when we might ask Google to find something for us and it auto-completes.
What happens when AI can do what Sherlock Holmes does in his mind palace, and then go off and solve crimes, bake the perfect soufflé, or even build the next Unicorn … on its own?
