If you, like most people (or, at least, like most people who think about things a bit, and often need to look things up), have found yourself using Wikipedia on a regular basis, you’ll have noticed that while it’s very good as a reference for uncontroversial topics, it is consistently left-biased wherever the subject matter touches anything subject to political, cultural, or ideological polarization (which, these days, is pretty much anything that isn’t chemistry, astrophysics, or mechanical engineering).
Elon Musk, true to form, was annoyed by this, and so he has introduced an alternative: Grokipedia. It’s just getting started, and has a long way to go — but as you’ll see from this refreshingly frank article on race and intelligence, it already shows real promise.
4 Comments
So Malcolm, to be clear for a Gen-X dinosaur like myself: did one of Musk’s AI entities just scour the Internet for everything (pro and con) on the veracity of HBD, and this is what “it” neutrally and objectively came up with? Just skimming the essay, it strikes me as accurate (or perhaps better, truthful or honest), but biased.
As it happens, I decided to look up Nick Fuentes’ entry on Grokipedia, because of its topicality due to Carlson’s recent interview of the guy. Again, skimming the article it strikes me as quite empirical but perhaps also slanted in the provocateur’s defense. This maybe suggests something: that even an all-knowing device like AI, able to be the perfect datanaut by finding every conceivable variable and putting it in a proper slot – will still arrive at an interpretation that is, well, personal. Which if it were ever to achieve omnipotence over humanity, that does not exactly auger well. To make an analogy – while one might admire the insights of a Chris Langan, whose brain power certainly exceeds my own, I sure as hell wouldn’t want this rather quirky fellow to have power over me.
Just one more thing Malcolm. Maybe I’m missing it, but is there an explanation anywhere on Grokipedia about the algorithms it’s using, how it finds and evaluates information? Kind of a moral and ethical thing to do for the reader Mr. Elon Musk, you Master of the Universe!
HI Jason,
Interesting points all. (Not sure what to make of “accurate but biased”; I suppose that means “not reporting falsehoods, but only reporting truths that support a preferred position”.)
I think “bias” might in many cases be just a somewhat pejorative way of expressing what a Bayesian might just call “updating one’s priors” — that is, revising one’s analyses and predictions as more data come in, and patterns and regularities become clearer. All honestly intelligent people do this, and it’s high time we stopped stigmatizing it.
There are (at the very least) two factors involved in what comes out of AI systems like Grok: how its training data are curated and filtered, and what it does inside the “black box”. The former, at least, could be made public so as to reassure Grokipedia users, but the latter is entirely mysterious, even to Grok’s engineers and handlers.