I think it speaks to the broader notion of AGI as well.
Claude is definitively trained on the process of coding not just the code, that much is clear.
Codex has the same limitation but not quite as bad.
This may be a result of Anthropic using 'user cues' with respect to what are good completions and not, and feeding that into the tuning, among other things.
Anthropic is winning coding and related tasks because they're focused on that, Google is probably oriented towards a more general solution, and so, it's stuck in 'jack of all trades master of none' mode.
Jack of all trades, master of none, is oftentimes better than master of one.
You wouldn't download a CAR, would you? You wouldn't hack your own INSULIN pump, would you?
Face it: If it's GPL and vulnerable to interference, responsibility is squarely on the manufacturer and the fastest death-free way to prove it. If it's GPL and modified by the owner, fuck off.
If everything is information, then nothing is.
A disordered system still has state. You just don't know what it is.
Around 2010, I was talking about this with friend about this failing of mine, and he said "you should try python, I've heard it is popular with non-math folks." So I bought a book, and as soon as I opened it, I could just read it. It took me a couple days of reading to wrap my head around object orientation, but on the functional side, I could have written fizz buzz like, maybe half an hour after opening the book.
Humans have logic pre-built into our brains, it's just that we use natural language as our syntax. Python cleverly used as much of the natural language syntax as was practicable to remove the barriers to entry for non-math majors. Whitespace is perfect example of a natural language syntax feature.
Remember that Apple SSL bug "goto fail"? That was a whitespace bug, because even if the C feature predated python, everyone's eyes had been trained to slide right off that particularly crass shortcut as python was widespread by that point.
So what if "everyone in Seattle hates AI"? What gives The Author the right to simultaneously invalidate Seattle's comparatively immeasurably larger advantage in experience, qualification, and education? If even the ludicrously biased title had even the barest hint of truth to it, they've stacked the deck against themselves in credibility unless they've already mentally biased themselves to blindly dismiss anyone that doesn't mirror their own now blatant fanaticism. Which we've already established now includes all of Seattle.
So put this out on the curb with the rest of the garbage meant to inflame and divide, because on it's face it is neither reasonable nor factual.
> Bertrand Russell, one of the great intellectuals of his generation, was known by most as the founder of analytic philosophy
That title is usually attributed to Gottlob Frege (in particular his 1884 book "Grundlagen der Arithmetik", and his 1892 paper "Über Sinn und Bedeutung") who directly influenced Bertrand Russell, Rudolph Carnap, and Ludwig Wittgenstein, who all later became large influences on analytic philosophy themselves. Frege is most known for the invention of modern predicate logic.
-- C. A. R. HoareThe author might think they're being a meta-pirate by saying "oh these are all possible actions proscribed by most organizations" implying that the proscription itself is a sign of ossified incompetence, instead of identifying the underlying one-true-savior fallacy that underpins it all.
The general danger here is human error. The point of leveraging a collaborative environment is to design process to detect and remediate human error before it radiates outward into more cost. The farther it goes, generally, the higher the cost, non-linearly. It shouldn't be "never do this", but instead "if you're going to do this use every tool at your disposal to make sure it's done correctly." Siloing the entire decision tree to yourself is exactly how not to do it.