Teen pregnancy rates are down since the mass adoption of the internet, a kid learning a few years early that there exist sexualities other than the default one will affect them much less than losing internet privacy and anonymity for life.
The great thing about AI coding is that you never have to learn enough judgment or context to know if something should be written a certain way or not!
Been trying to think it through, but I guess I'm getting annoyed by how trivial people seem to think fines are.
I'm not sure how to research this, but it seems like most companies hit with a fine like that will change their behavior. Malicious compliance exists (and I would not put it past Google), but it is at least moving in a better direction.
That said. I also think Google has access to powerful tools to manipulate the laws behind the fines....
Google just got hit with a $314 million fine for secretly tracking Android users’ locations. Sounds big, right? Not really. In 2024, Alphabet made $62 billion in net income — this fine is just 0.5% of that.
History shows fines this small don’t change corporate behavior.
Take HSBC. In 2012, they paid $1.9 billion for laundering cartel and sanctioned-country money — 11% of their profit that year. Still, they were later linked to another $4.2 billion laundering ring between 2014–2017[1][2].
Or Pfizer. In 2009, they were fined $2.3 billion for illegally marketing drugs — about 25% of that year’s profit. Yet more settlements followed in the years after[3][4].
If 10–25% fines didn’t deter repeat offenses, a 0.5% fine won’t even register. Google will just move on and likely continue the same behavior.
People here think the fine is trivial because it is. Unless you include some sort of regulatory oversight or criminal charges, corporate behavior doesn't change, it's just the cost of doing business to them.
> Google just got hit with a $314 million fine for secretly tracking Android users’ locations
It didn't. It got sued for using users' metered cellular data and received a judgment telling it to pay users for that data usage. There is nothing about a fine for secretly tracking locations here.
"Bias" and "narrative" have become meaningless words that people use when encountering ideas they don't like. Or when they believe that neither side could ever be right about an issue, that it's virtuous for a stance to be in the middle of the current Overton window.
Even presenting a given problem as "merely" two-sided is often disingenuous.
For instance, if you "just" look at abortion, trans rights, and immigration, you may assume the two parties in america present diametrically-opposed groups of people (...which is even itself quite debatable). But this is only because the two parties don't differ much (or at least, don't bother to platform enough to evaluate) on most topics politics could be about.
I don't think it's entirely unreasonable to presume that more valid positions are likely to be found closer to the midpoint of the current Overton window. I suspect that as a matter of Bayesian probability, more extreme positions are more likely to be wrong.
The whole premise of the Overton window is that it does not represent the full spectrum of opinions on some issues, which is unchanging, but rather some socially constructed window onto that spectrum, which does change. Assuming that the midpoint is probably correct is equivalent to saying that any change from the status quo is probably bad. Is that what you really mean?
> The whole premise of the Overton window is that it does not represent the full spectrum of opinions on some issues, which is unchanging, but rather some socially constructed window onto that spectrum, which does change.
Why would you assume that the full spectrum of opinions on an issue would be unchanging?
The concept of the Overton window is meant to represents how attitudes do change in response to variations in perceived extremes.
> Assuming that the midpoint is probably correct is equivalent to saying that any change from the status quo is probably bad. Is that what you really mean?
Basically, yes. I you have a mature system that's reached a stable equilibrium, then its parameters are likely to be optimized against its inputs and constraints, so deviation from that equilibrium is likely to produce less optimal results. If the inputs or constraints have changed significantly, of course, then that might not be applicable.
> Why would you assume that the full spectrum of opinions on an issue would be unchanging?
The spectrum of actual opinions (which is related to the Overton window) may not be unchanging, but the spectrum of all possible opinions by definition can't change. The range of opinions on nuclear weapons goes from "Let's use all of the nukes that exist right now, and also build more and use those, at any cost." to "Nuclear weapons are so evil that any amount of sacrifice is justifiable if it effects even a minuscule change in the likelihood of nuclear weapons being used, or more being built, or existing ones dismantled." This has been true since they were invented, and will continue to be true forever. But both opinions are outside the Overton window, which changes over time.
Based on a gander at history, I doubt it. Usually one side is actually right and it seems very obvious now but was not obvious at the time. And, typically, the moderate or midpoint position seem incredibly stupid now. Like the 3/5ths compromise.
While I'd favor defining words rather than claiming they're meaningless, those aren't accurate definitions.
"Narrative" in this context is more like "attempting to steer the current discourse by making connections among things and positioning them as part of a coherent story", which can be positive (if the result is accurate and reasonable and helps people better understand what's going on) or negative (if it's spin or manufacturing consent). It's "narrative" to say "you should be afraid of X, it's the cause of all your problems". It's also "narrative" to say "here are the five different things we're currently doing to improve Y, and how they tie together into a coherent picture".
Also, bias would be easier to deal with if it were always "lying", or if it advanced a coherent position. It's much broader than that.
Why not? Current open models are more capable than the best models from 6 months back. You have a choice to use a model that is 6 months old - if you still choose to use the closed version that’s on you.
It's wild that people don't see that LLMs are following the same playbook as streaming etc. and in time will predatorily monetize in every way possible. If you think people are trapped as customers because they can't do without tv shows, imagine five years from now when it's general thinking that people have become dependent on the tech giants for.
