In case you haven’t been following this trend, artificial intelligence has moved well beyond the plotline of the next Terminator sequel.
In actuality, AI side-stepped science fiction accounts with killer robots many years ago. Lately, it has entered into a highly sensitive area which really matters. Really! By that I mean it matters to me, personally, directly, (and what could be more important?), but also maybe some other reporters and wordsmiths in a variety of trades-- like presidential speech writers.
Could some AI system with a quirky name take my job and livelihood away from me any time soon? Or could AI cause my old writing professors to stop teaching how to create pithy headlines that get the most page views when a relatively dumb (but expensive and shiny) machine will someday soon do it much better?
One recent bit of language model research from OpenAI knocked me sideways.
This relatively new GPT-3 model, when properly set up, can write creative fiction. (That means, I guess, better fiction than you usually see on Twitter.) The writing samples have been described as “not just close to a human level but creative, witty, and often beautiful.”
Dang. I think an editor once called something I wrote witty, but “creative, witty and beautiful” in one full stroke? Never.
I love this one account of how the GPT-3 model talked via text to a human. This is an actual AI encounter. I am not making this up!
Human: Tell me a story about the time the Cat stole the sun.
AI: Okay, let’s begin. First you must know that the sun is actually a cat.
(Right off, you can see that this is similar to a conversation between a very tired human dad putting his little daughter to bed and infinitely more creative than anything the dad encountered at his work in weeks. But later the conversation continues…)
AI: Ok, the sun used to live at the top of a green mountain. One day a hungry cat came to the mountain and climbed up to get the sun. But when the cat got to the top, there was no sun there. There was only a hole, that went straight to the sky. (Did you see that? The stupid AI still hasn’t mastered commas.) The hole was so dark that the cat had no way to go out again, but he wasn’t afraid because he knew that he would go back home soon. So, the cat climbed down the mountain and took some of the sun with him.
To my great relief, GPT-3 still lacks common sense and the ability to think abstractly (although I’d argue a cat stealing the sun qualifies as abstract). Those AI deficiencies keep me miles ahead (on my good days) although I do hear the footsteps behind me getting closer and closer in a very dark alley.
RELATED: Microsoft Azure extols virtues of Nvidia A100 chips for AI work
Headless body, topless bar
One of my earliest memories of writing about AI for a different publication happened when I was sitting at my desk looking bored and my manager asked me to cover a neuromorphic computing conference in downtown Boston. I went after googling “neuromorphic.” It was a rainy day. Many of the conference presenters wore tweed jackets and had long, unkempt hair or were bald with long beards. (Improving their credibility of course.)
One especially ancient man projected on a screen various AI samples of random headlines matching English subjects and verbs: “Man-flies, Car-stoops, Clown-vomits.” You get the idea…It was an early light-hearted example of how a machine could at least employ words in proper syntax, even if they weren’t exactly logical for meaning. Some headlines were logical and others weren’t, by chance, but a lot more was happening than was missing in his demonstration. He also talked at length about neurons in the human brain and how far machines had to go to reach human networking capability. That was ten years ago.
I remember leaving and thinking that I had a duty to describe for my publication what he and others said at the event without sounding sarcastic, a nasty side-effect of working in journalism. I also felt pretty sure it would take a long, long time for an AI model to approximate my unique level of sarcasm, much less Jerry Seinfeld’s or, say, Henny Youngman’s. Seriously, how long will it take for a machine to be able produce this headline gem: “Headless body in topless bar” penned by Vincent Musetto, a human, of The New York Post on April 15, 1983.
Between that event in Boston a decade ago and today, a phenomenon called automated journalism has emerged, sometimes called robot journalism. News articles are generated by computers through AI software. In May, Microsoft replaced dozens of its MSN contract journalists with robots. Reportedly, their duties include optimizing content for better page views by rewriting headlines from stories received from publishing partners.
To my relief, computational journalism professor Nicholas Diakopoulous at Northwestern University has downplayed the idea that bots will replace journalists. His research estimates that only 15% of a reporter’s job and 9% of an editor’s job could be automated..
Agate monkeys
His finding makes sense based on my early memories of writing for an old-fashioned community newspaper, now defunct, that covered the things that really matter-- like high school swim competitions. Parents wanted to see their kid’s name in print (yes, in print) and the paper would hire out a small army of young reporters to take phone calls for a few hours after a series of county-wide swim meets between a zillion high schools. Coaches would report each swimmer’s name, event, time and place of finish over the phone. In any given swim meet, there were gazillions of swimmers! On any given Saturday, there were more swim meets than humanly imaginable!
That information had to be typed up, then submitted into an early form of content management system, and printed on a page in agate type, a particularly cruel measure with 14 tiny lines of copy per column inch. Basically, it meant that grandparents could not read without glasses what their swim stars had done. They called the reporters taking the stats agate monkeys. I was an agate monkey exactly one time because I needed the money. I never wanted to be an agate monkey again. I had a master’s degree, for pete’s sake.
It’s easy to see how a community newspaper today (there aren’t many left) could create a smartphone app asking coaches or the water boy to input event, time and other info that would almost instantly, wirelessly be plopped onto a web page. The only challenge would be making sure every coach or team volunteer actually input the data or kept the smartphone charged. No more agate monkeys!
My point in telling this story is that clearly there are many emerging roles for automation with AI in all kinds of professions and that AI is getting, yes, more creative. This capability will streamline the busy-work jobs that allow workers to focus on higher level tasks. Yes, I’m sure you’ve heard this line of thinking before. We and our children will all be doing more dignified and intellectually-challenging work in the future, thanks to AI. And doing it ever-more creatively! (We still have to figure out how people will be paid, of course.)
Seriously, there’s much more coming from AI that’s positive, outweighing the negative. Believe it!
AI will liberate us from burdensome tasks so we have more time to hunt down the killer robots.
RELATED: Misty the robot graces Fierce AI Week in engineering
Matt Hamblen is editor of Fierce Electronics.