Very true, but the benefit to one-on-one instruction is so enormous that we should find ways to apply it fractionally if we can’t apply it fully. One thinks eg of the one-room schoolhouses of the 1800s, with younger students learning from older students.
Writing SQL, I'll give ChatGPT the schema for 5 different tables. It habitually generates solutions with columns that don't exist. So, naturally, I append, "By the way, TableA has no column FieldB." Then it just imagines a different one. Or, I'll say, "Do not generate a solution with any table-col pair not provided above." It doesn't listen to that at all.
You do understand that these models are not sentient and are subject to hundreds of internal prompts, weights, and a training set right?
They can’t generate knowledge that isn’t in their corpus and the act of prompting (yes, even with agents ffs) is more akin to playing pachinko than it is pool?
This is something that people working on extremely simple apps don’t understand because for their purposes it looks like magic.
If you know what you’re doing and you’re trying to achieve something other than the same tutorials that have been pasted all over the internet the non-deterministic pattern machine is going to generate plausible bs.
They’ll tell you any number of things that you’re supposedly doing wrong without understanding what the machine is actually doing under the hood.
It's funny, we've watched for two decades as the click-driven dynamics of the internet have degraded the meanings of words. At first, I was outraged on a daily basis. Then, as we all did, I learned, against my will, to forgive. "Can't blame them for chasing clicks! Who among us wouldn't cheapen a word if it meant a view?"
But - and this is the funny part - I feel like my teen-angsty self has been vindicated. I'm so burnt out on exaggeration, not a single news site has gotten regular clicks from me in over a decade, nor do I comment or read comments. I listen to a little history dork YouTube before bed, or for tutorials. I'm free.
Haven't listened to Librivox in years and years, but I still fork over the annual $2.99 because I feel I owe it.
It's horizon-broadening. Lots and lots of interesting reads/listens I never would've picked up otherwise. 1800s ghost stories, darkly racist novels like The Leopard's Spots (good luck getting through the first 10 pages). My favorite is Havelock the Dane: A Tale of Old Grimsby, first written circa the 14th century but thought to be much older. When you listen to it, it is apparent that the author and the intended audience know 100x more about nautical things than you do. It's also charmingly simplistic; the main character is sort of like Conan the Barbarian. He'll do things like "lift a stone the weight of an ox and throw it the length of two men." You imagine the audience being like, "Oh my fucking god.... that's amazing."
In my time in management, I found that the commonplace psychological descriptors we use failed to adequately describe what I was seeing. Two employees may both be “detail-oriented,” but there are subcategories within that depending on where the motivation comes from and those subcategories behave differently. Some people want that A+, some people like their squares square and their circles round (and it gives them anxiety when they’re not). Those groups are different, and I don’t know what to call them.
At bottom, there are ‘simple machines’ of psychology that, in combination with each other, produce behavioral traits at the top level. We don’t really have words for them, or at least not words I can think of for the ones I see.
