Prompt
Well, this probably went about as well as it ever could. AI text generators
like Inferkit have a long way to go before they’re even moderately
intelligible, but the experiment is over for now.
Botner
My failure shows that the sort of interface to natural language processing
researchers were designing has a long way to go. Fortunately, it’s something
we’re only slowly developing. No matter. I have a few more strings to write.
UPDATE: A final comment from the Inferkit folks. Now, more than three years
after we created the text generator, a developer working on an AI software
engineer team (otherwise known as an AI “computer”) actually leveraged the
text generator to understand the task: Thanks! It’s neat to see your
software applied to AI instead of computer science. I’m looking forward to
reading your latest work... In the meantime, what’s in the future for AI
software engineers? I’m glad you asked. We’re also building an AI software
engineer training curriculum on Coursera. P.S. It turns out our input
generator makes sense when you write in English. Quotes and inspiration from
AI researchers: Want to read more? Check out my book, Everybody Writes In
Hieroglyphics.
Conclusion
See what I mean? Like, what was that? It wasn’t my failure, it was the
system’s. I did everything I could. That’s the problem with this text
generator, and when I had to use a different once, because Inferkit was
down, I encountered the same problem. You’ll notice I often very
obviously—to a human at least—tried to send the story in one direction, and
the AI would do the opposite. There was one instance where I said that I
hated soup, and so the AI decided to write a story about how much I loved
the soup that I made. Well, that doesn’t make any sense. All it has seemed
to ever do is pick out keywords, and make something up. It doesn’t appear to
comprehend context or true meaning. The prompts weren’t all that relevant to
it. I mentioned soup, so it came up with something about soup, using other
keywords that are traditionally associated with soup, but which don’t
necessarily have anything to do with what I said about soup. That’s fine,
this was an experiment, and I intend to do this again, maybe in ten years,
to see how far the technology has come. Hopefully, it will indeed have
improved, and if it hasn’t by that point, I would strongly suggest you
invest in silver, and build an underground bunker, because civilization is
probably doomed. We won’t make it past our Great Filter if our artificial
intelligence can’t even recognize that the snow in Texas is bad, and that I
don’t want God to give us more of it. Anyway, I don’t regret this series,
but I am kind of glad it’s over. I’ll be getting back to my canon after
this, for a series that takes place in multiple universes (as many as
99—we’ll see). For now, say goodbye to Botner. I’m shutting him down,
sticking him in a storage closet, and moving on. I guess that’s the great
thing about robots. Hopefully he doesn’t rise up against me because of it,
though.
No comments :
Post a Comment