How much do you really know about AI - as a system? There are lots of moving parts to the system. Right now I am going to focus on a mere fraction of the AI system: the knowledge base. This particular “chunk” of the system is extremely important; and what is any intelligent system without a treasure trove of knowledge, right?
Let’s go back, way back, to a time before the World Wide Web was freakin’ everywhere. Back in those days, most of us in tech were using Emacs — ok, a bit of my personal tool zealotry is leaking out here — and we had access to a dynamic source of knowledge across the globe via ‘rec dot’ or some form of BBS (Bulletin Board System). The rest of the world was reading a more static source of knowledge via newspaper articles, watching the news, or listening to the radio. For just about everyone, information was difficult to find because you couldn’t just quickly execute a random fit of searches in some “app”. You had to KNOW about someone’s writing to benefit from their perspective. Either a friend would send you an email with a link or maybe you’d be talking about it via chat (rare) or sitting with one another drinking a coffee or over a meal (more frequent). Now, fast forward to today…
How fast, and easy, is it to search for something you would like to know? What repository of information is the “well” from which those results are drawn? And, are those results “pure”? How often do you need to apply your personally crafted filter(s) to vet the information you are reading? Finally, is all the information static or is it… malleable? Can you honestly trust these sources?
I use malleable because the worst repository is a Wiki. Anyone can post, anyone can edit, and anyone can remove a post or an edit, therefore if your source is a Wiki it in all honesty is a less than optimum repository. Additionally, much of what is posted on the internet (even my writing) is subject to a set of deterministic filters, that quite often are dissimilar to yours, which “present” you with “results.” Because you now know this happens, you’ve got to hone your own filtration to balance the invisible yet palpable set of filters imposed upon you. A Wiki is curated and so to is the knowledge base fueling AI.
Argue all you want, but, it’s true… sorry.
What you need to come to terms with is that there are a set of biases within AI from the creators and there are also a set of biases applied to the knowledge base by the curators. I posit while the bias in the code / architecture of AI is bad, the bias in the knowledge base is the “square wheel” on the AI vehicle. Your AI engine will get to the answer in microseconds from petaflops of churning, but the result is a Chevy Citation or Ford Pinto! Civilization is gonna get rear-ended and we certainly DO NOT wanna be riding in AI’s version of a Ford Pinto, right?!
The next time you fire up the latest curated AI app think twice and ask yourself, “How were these results generated?”