The anti-anti-racist right is proceeding to build conspiracy theories around the new machine-learning services provided by big tech. This conspiracy theory seems to have more purchase among the ‘respectable’ conservatives than many others have (though frankly, multiple conspiracy theories are being parroted across the conservative media network). As usual, these conspiracy theories are meant to simultaneously stoke unwarranted fear against any influential person who isn’t a card-carrying conservative, while also distracting from the real problems surrounding an issue.
The reason I raise this warning here is a recently published essay in the Atlantic called “Why We Must Resist AI’s Soft Mind Control”*. It’s by some little known conservative hack with a few publications in places like the National Review (but he’s anti-Trump, which seems to be the only demand the Atlantic has for conservative writers these days). Unfortunately, it’s pay-walled, but there’s really nothing passing for a coherent argument in the essay, so that’s not a big loss. It’s typical of the “anti-woke” hand-wringing over the output of the Google Gemini system (you may have heard about the multi-racial George Washingtons).
So anyway, here’s what I see as the logical errors and omissions:
1. These conspiracy theorists immediately assume that Google is pushing ideological propaganda, rather than trying to grab market share.
2. They assume that Gemini is a general-purpose tool, rather than something calibrated to serve a niche market. Google may present Gemini as a general-purpose tool (I don’t know), but that doesn’t mean that it is. Marketers routinely over-promise what they can deliver. Tech companies regularly sacrifice general-usability in order to optimize features for their primary customer base, but they never say so. To do this optimization, they often make assumptions about what the user wants, which seems to be exactly what Google is guilty of. For instance, Excel is optimized for use in business applications, at the expense of its use in scientific applications — yet Microsoft would never tell customers not to use Excel for scientific applications (see figure at right; but they apparently fixed the date issue recently).
3. The ‘woke AI’ conspiracy theorists assume that people are turning to the automated content generators as sources of information, when no reasonable person views them that way (although, again, Big Tech oversells them as such). The problem here is that some users may be taking the output of these systems seriously — rather than treating them as semi-random raw material might be shaped into something usable. It remains to be seen if these automated content generators have any real value other than as toys and making marketing fluff. If conservative hacks see these automated content generators as producing propaganda, it’s probably because they think that’s all that words are good for.
Throughout it all, the author criticized Gemini for ‘inconsistent reasoning’ — as if it reasons at all. Of course, by treating this automated content generators as a devious super-intelligence (an idea invoked by the term AI itself), the author is just contributing to the sales hype of the companies that are setting up for one of the biggest loss-leader/lock-in scams in history. After all, the only reason we have to worry about our society being ruled by soulless machines like Gemini is that we’ve already handed over control to a different type of soulless machines — we call them ‘corporations’.
*(edit) ‘Soft’ has become the favorite weasel word of pseudo-intellectual right wingers. They use it to provide deniability when they try to stick a scary label on something where it simply doesn’t fit. The other widespread use is “soft totalitarianism” — which has none of the scary traits of real totalitarianism. In other words, it’s just not totalitarianism.