The Case for Governance: Why AI is going to make us all managers
Google debuted their ChatGPT competitor "Bard" last week -- by all accounts before the company wanted to (and was ready for). The press release for it offered a little science fact - did you know that the James Webb Space Telescope was the first to take pictures of a planet outside our solar system? Well, it wasn't, and it happened 20 years before JWST even launched.
Then Microsoft added ChatGPT to Bing and it didn't take long before bad internet people made that robot feel real bad itself and badgered it into ignoring its prime directives to not be a jerk, and giving out all sorts of goofy code names and behavior laws Asimov would have been excited about.
Oh and hey, remember when AI made a never-ending Seinfeld episode that was streamed on Twitch? Until it made a series of transphobic jokes and got banned?
Wait.
This is the future?
Call in HR, because these employees need to be put on a performance improvement plan.
The knee-jerk reaction would be to say that this burgeoning technology is fundamentally flawed and it'll never do what humans can do. But the sad truth is that it is doing what humans do, just not what us "managers" want it to do. It's been trained on our best and our worst and everything in-between, and it spits it out like a first day intern with a lot of preconceived notions and unfortunate biases. That's a problem, but it is a problem solvable by traditional means the content world is familiar with: editorial review, workflow approval, content guidelines -- in short, governance.
To refer to another law written by an author from the golden age of science fiction, Theodore Sturgeon:
ninety percent of everything is crap
So by that metric, maybe AI isn't doing so bad. But when it does do bad, it goes all the way.
In the fight over the nascent AI competition for best of breed tools, the opening salvo has been around capabilities... how fast Dall-E can iterate over Midjourney to make fingers that look human, how much better Bard can be at answering general topical information over ChatGPT so search can be completely transformed, and which engine someone would actually trust to write some basic code (how are you going to know how bad it is without it getting down voted on StackExchange?!??!?).
But the future battle will be which tools provide the best tools for winnowing the wheat from the (quite possibly racist) chaff. And if I had to guess, those will be leaning just as heavily on AI as the systems those tools are meant to manage.
