Tuesday, July 19, 2022

startling tech

I was thinking about what technologies really startled me, in a "they an do that now?" sense.

A very incomplete list, in rough chronological order:
  • Google. It snuck up on me as the fallback search for Yahoo back in the day. When I realized it did a much better job of searching my site than my own homebrew search functions... and it was doing that for, like, every site on the web?
  • iPod. Again, a matter of scale: 1,000 songs in your pocket is just amazing.
  • Youtube. Yet another bit of astonishment at the scale of it. Hosting that much video for free was just mind blowing.
  • Google Maps. This was a minor thing, but the navigation it provided in browser, these large seamless draggable and zoomable tiles... amazing
  • On-dash GPS. I remember going to the video game con "PhillyClassic" and the guy driving had one of these. What a miracle of making life easier.
  • Accurate Video Game Emulation - especially in-browser.
  • AlphaZero. I always said I would be impressed when a good chess playing program was also good at another game - like it formed its own intelligence about a game, as AlphaZero did with Go and Chess. (Now the bar is "when will the computer be BORED of playing chess")
  • GPT-3 etc text generation. The "Ghostwriter" segment of The Ghost in the Machine episode of This American Life just blew me away... and you can find instances of it online. The ability to generate text that seems to have a point of view.
  • DALL-E etc art generation. Boy O Boy. Making illustrations from arbitary text prompts...
Those last 3... man.

One model of intelligence is that it's all metaphors and connections. And computers are really getting there.

I think the future of creative and information based jobs will be learning how to harness these kind of forces. And that's a little scary- both from a "future of work" standpoint, and also because so far I'm not sure if the machines will get better at showing their work, explaining their reasoning... (and right now they sometimes exhibit the racism etc implicit in their training data sets...)




No comments:

Post a Comment