Bot authors a most-read article at The Next Web

The journalists at The Next Web scripted a bot — Satoshi Nakaboto — that crawls around the aether looking for Bitcoin news and tweets, then assembles into a daily news story. There's no cutting-edge AI at hand here — it's just good-old "rule-based phrases and terms we wrote beforehand," as The Next Web folks note. You can see an archive of the bot's daily stories here.

Nonetheless, it wasn't long before the bot published a story that was the #1 most-read for the site that day, as the editor noted:

Tweet by editor of The Next Web

Már Másson Maack, a writer for the site, wrote a good essay pondering his own psychological reactions to being outperformed at his job by a few lines of code. As he notes, there's an obvious discomfort at realizing a bot can author a top-viewed story, because of course in a publishing operation, money comes from reader clicks; if a deathless, tireless, un-unionizable bot can deliver those clicks, that's unsettling for all the meatbag authors.

As Másson Maack points out, the traditional response of AI optimists is to argue that this will all be fine. Sure, bots can outperform humans at routine drudge work. But this frees the humans up to do the complex, thoughtful, high-EQ stuff that humans are at the moment uniquely capable of, right?

Fair enough, in theory. Except Másson Maack identifies the bigger problem: Our economy mostly has no idea how to value those supposedly superior "human" skills.

No, today's firms are mostly obsessed with measuring output, which is where bots excel:

The phrase du jour of technologists, 'tech-minded' CEOs, and other self-proclaimed thinkfluencers is that AI will actually make 'work more human.' The argument being that with the rise of AI, 'soft skills' like communication, creativity, teamwork, and problem solving will become the most valuable talents in the future job market.

But the truth is we've never fully known how to evaluate those skills — even before the Satoshi Nakabotos of the world started popping up. The struggle with finding a quantitative measure for journalism also applies to soft skills in general. How do we know if someone is great at problem solving? Maybe when their department manages to surpass their earnings target?

All the measurements we've come up with for job performances are incredibly rigid and don't account for the multiplicity of the human experience — but great for bots. And even though you might get the freedom to create something cool, the fact is that everybody answers to somebody. What about the goals and responsibilities of your boss and your boss' boss? What about the shareholders? What about the OKRs?!

Nailed it.

(CC-2.0-licensed photo of the pencil-carrying bot courtesy the Flickr stream of Matthew Hurst')