Dem
47
GOP
53
Most people outside of the city of Los Angeles had never heard of Patrick Soon-Shiong. But then he beat Jeff Bezos to the punch in selling out his newspaper, The Los Angeles Times, to Donald Trump. That made him something of a national figure, and not in a good way.
Still, Soon-Shiong has pressed on, and has been very enthusiastic about two notions in particular. The first is "balance" and the second is using AI to somehow help improve the content of the paper. Both concepts should make anyone who knows anything about journalism run screaming. The only way it could get worse is if they are somehow paired.
That is exactly what happened early this week, as the newspaper—on Soon-Shiong's instigation—unleashed its new toy (provided by the company Perplexity) to provide "additional perspectives" for pieces in its opinion section. One of the op-eds that the AI was asked to comment on was a piece by columnist Gustavo Arellano, about how this year was the 100th anniversary of when the Ku Klux Klan took control of the city council of Anaheim, and how the current city council should have done something to acknowledge the (fairly brief) reign of the hate group. At this point, you can probably see where this is going, not unlike a car crash that cannot be stopped. The AI, in service of "competing" perspectives, mounted a semi-defense of the KKK:
Local historical accounts occasionally frame the 1920s Klan as a product of "white Protestant culture" responding to societal changes rather than an explicitly hate-driven movement, minimizing its ideological threat.
Oops.
Now, as chance would have it, this is a subject that (Z) knows an awful lot about. That is because the very first class he took in grad school was an urban history course with the late, great Eric Monkkonen, and the final paper we wrote for that course was about... the KKK's seizure of political power in Anaheim in the 1920s. And the AI is actually vaguely in the ballpark in terms of two different, potentially helpful, points here. The first is that the KKK has not existed for all of the past 150 years, and has had three different iterations (1860s-70s, 1910s-20s, and 1950s-present). And the second is that the gap between the racism of the 1920s Klansmen and the racism of non-1920s Klansmen was not as large as the gaps were in the first or third eras of the KKK. That is due, in part, to the fact that most everyone in the 1920s was pretty racist, and in part to the fact that the 1920s KKK was not driven solely by racism, but also by xenophobia and antisemitism. It was still a hate-driven group, of course, it's just that the dynamics were somewhat different than with KKK v1.0 and KKK v3.0.
Anyhow, you can kind of see how AI might review the scholarship and come up with a statement like the one it came up with. At the same time, you can also see how AI is very bad at nuance and subtlety, and also in knowing which circumstances require extra amounts of nuance and subtlety. The notion is that one day, presumably soon, it will get better, but we're not so sure. (Z)'s phone has been giving driving directions for 10 years, and still mispronounces half the street names in Los Angeles. For roughly that same amount of time, (Z) has been dictating messages to the phone, and it still can't get "Did Otto potty when you walked him?" correct, invariably rendering it as "Did auto-party when you walked him?" We don't even know what auto-party is. Maybe it's what AI does when it's off the clock. In any case, if the computers can't grasp "auto" vs. "Otto" or figure out how to pronounce "Sepulveda" or "Cahuenga" after a decade, how can they possibly produce intelligent commentary anytime soon? Or ever?
It would seem that someone at The Los Angeles Times has figured this out. Maybe it was Soon-Shiong, as he was busy wiping egg off his face, or maybe it was someone else. In any case, after just one day in use, the AI commentary has been yanked, at least for now. (Z)