Much like the first Christmas itself, the biggest story of 2022 is not the headline-grabbing collapse of crypto, FTX, Sam Bankman-Fried, bonds, stocks, or gas prices; it's the rather quieter but much more significant arrival of ChatGPT.
Case in point: on Sunday evening, two of our family members had to quickly write a last-minute, rhyming scavenger hunt for their nieces to do the next day. For both of them, this was about as fun as a root canal. They came to me for writing help; I was all excited to pitch in, but also in the middle of managing the chaos of giving the kids dinner, and thus a little delayed.
No matter; my husband sprang forth with an idea. "Just use Playground!" He said. They looked at him blankly. Playground, which is just a different interface of OpenAI's ChatGPT technology, is a website that lets you enter anything you need and get an AI-generated response. These technologies have absolutely taken off lately, and not just for text; it all started with image generation (Dall-E and Stable Diffusion), and now you can try it with music, and I can only imagine where it's going from here.
So for instance, one of the clues in the scavenger hunt was going to be hidden behind a bench in the family's laundry room. My husband typed the following prompt into Playground: "Tell me a rhyme about the bench in the laundry room, without using the words bench or laundry." And back came the following reply:
A place to sit, so nice and sweet
A spot for folding, a place to meet
Where warm clothes lie, from washing machine,
A place for comfort, a household scene
A stunned silence filled the air. A slow smile spread across my sister-in-law's face. You could see the gleam in her eyes; freedom!!! it clearly conveyed.
Now, Playground's answers weren't always that spot-on for every clue. It actually got rather tedious after a few tries of trying to get each rhyme right, and I'm happy to say I sailed in later to the rescue to quickly finish off the project. But they could have done it without me. Just like people will now be able to write resumes that are basically good enough, generate email replies that are useful enough, and so on. The news is already full of stories exploring cheating and plagiarizing with this technology.
If you think Facebook's had it tough trying to moderate usage of its platform--or Twitter, for that matter--imagine the moderation demands that AI will require. A bunch of programmers is not the societally responsible choice to put in charge of such a task. Something more like the editorial boards of the Wall Street Journal or New York Times is required. And yes, there probably will be--and should be--dueling political agendas between future platforms, because let's not pretend there is one "perfect" answer to things like the usage of fossil fuels.
But "ChatGPT" and all such emerging AI is about so much more than all of that. People have already used it to code websites. It's kind of like having a superpowered Google or Siri--a personal assistant who can do anything--and it's no surprise that Google and Apple are seen as two potential competitors. We are quickly moving to a command-driven internet that executes functions, where users will expect the technology to know how to give them what they want, instead of having to hunt down answers themselves for everything.
One final interesting caveat: ChatGPT's "answers" are based on essentially the internet up through 2021. This will come to be known as the "good old days" before AI started spitting out what critics might call "gobbeldygook" responses to all of our queries (ask it a basic question, like "Who is [your spouse's name]?" and unless they're a Wikipedia-quality celebrity you'll see what I mean. Some experts now worry these nonsense answers will pollute future updates, making the AI less intelligent in the future.
We can only hope. Because it's otherwise hard to see how this will not go absolutely mainstream and re-upend society as we know it in a very short period of time.
See you at 1 p.m!
Kelly
Tidak ada komentar:
Posting Komentar