You know, it would be interesting to apply this to the whole company. Want to vertically integrate everything? Sure, but be ready for any infraction to bring everything to a halt.
Do you know that strongly triggering these regions at the same time/repeatedly/in various patterns doesn't have any effect on your mood, wants, needs and so on due to experience in the field? Or why are you just discounting the idea?
IMO it's terrifying to imagine that adtech companies could plug this stuff into their optimizers and just run an experiment on broad swaths of the population to figure out if there's any effects.
It's a silly distinction. Unless we're wiped out by a virus or something similar, whatever ends us is likely to also take down a huge part of the biosphere as well - as is already happening.
"Burning something down" doesn't mean literally killing all life. But the unnecessary deaths of likely billions of animals in the long run should weigh enough to stop this nitpicking.
While I'm not sure I agree that the quote is applicable here, it is generally an incredibly useful way to analyze systems.
Reality doesn't care what you're hoping to achieve with a system, the outcome won't change based on thoughts and ideas. Since reality tends to be fairly complicated, we usually won't design perfect systems on the first try - instead we have to analyze the outcome, and make changes to nudge the results into the direction we want.
But what if we don't do that? What if we stick with the initial approach even after we measure that the outcome doesn't align with our initial intentions, maybe even introduces unwanted side effects?
It would be fair to say "well, we tried our best" if nobody was able to have hidden intentions, but that's not the world we live in. People can be very, very happy about our systems not aligning with our intentions & introducing unwanted side effects because this misalignment advantages them in some way. These people can then block changes to the system to keep these advantages alive.
I believe a small fraction of decision makers is enough to keep most systems fixed in such a misaligned state, and we'd never do anything about it if we followed your logic.
> They are much faster to internalize the time but slower to convert to numbers.
People keep saying that, but even after trying all my life, I still can't read analog clocks without mentally "decoding" the exact time. And I'm usually good at quickly building visual intuitions!
Though I think I experience what you're describing, just with digital clocks. For those it feels like I'm not reading 4 separate digits, as if the general "shape" is enough.
You say that as if inexperienced devs would naturally drift towards doing data validation without TS, but that hasn't been my experience. Instead what I used to see were lots of unnecessary/repeated checks spread throughout most source files (and which frequently didn't catch the actually problematic cases), which I've seen much less frequently in TS projects.
Even without TS, I'd be using zod (or a similar library). I haven't had any issues with zod's bundle size, and I don't see a good reason to write custom encoders/decoders everywhere.
How does TS make these situations worse? Seems to me that it helps a ton with identifying and fixing these issues. Either you can get away with just replacing the assertions with type guards, or you'll have to refactor some stuff (which is also much easier with types).
I've been really, really happy with oxlint. It has all the rules I usually need and the configs etc. tend to just work, whereas I don't know how much time I spent getting ESLint to work in slightly more complex repo setups.
> Considering that you can’t fire people in Germany for taking an excessive number of sick days (even if you have good reason to believe they are faking it that doesn’t seem unreasonable)
This is not true, it is possible to fire people in these cases.
Six months is the default probationary period in Germany, I can't remember ever seeing a job without that.
Temporary contracts are also a thing here, but if there's no objective reason for the contract to be temporary it will end after max. two years. According to Verdi ~1/13 contracts are temporary - not great, but could be much worse.
According to current laws, using LLMs to plagiarize existing works is not illegal.
There is no precedent because the technology hasn't existed before. There's tons upon tons of precedent for companies plagiarizing the works of small artists, and the justice system sides with large companies far more often than small artists. But none of the technologies you've listed allowed companies to automatically plagiarize the works of small artists, whereas LLMs do allow this.
There is a fundamental difference between those technologies and generative AI. I ask again: what stops companies from creating literally dozens, hundreds or thousands of copies of any creative work you publish? Why shouldn't they create AI artists that copy small artists in a "legally distinct" manner, and at a volume that drowns them out?
As these models get better and better, the only limiting factor is the available compute. Meanwhile data centers are being built like there's no tomorrow.
And now discovery is becoming exponentially harder, because every niche can get flooded with AI slop by companies trying to extract profits from real people's creativity.
Even if you use these tools to create something amazing, what keeps hundreds of variations "inspired by, but totally not copied" of your creative work from popping up? As these models get cheaper and more powerful, this issue will only get worse and worse.
I just don't understand how the multi-billion dollar industries (who employ hordes of people whose entire careers are spent researching how to best influence the average person) keep winning!
Only for non-erasable Typescript! The support for running erasable TS is here to stay. It's not ideal, but on the other hand this is paving the way for the TC39 proposal - if that ever progresses...