If you are an inveterate optimizer, it works to your advantage. If you are not, well I guess you just don't care. So net net, yes we prefer this equilibrium.
I think there's a fourth class (to the third class another commenter mentions) that you're missing here, which the credit card example makes obvious: people who fall for the "human behavior" trick the company aims for, and are exploited or hurt by those practices.
Credit card case: people prone to overspending and falling in a debt loop, lured in by $500 due to being poor or in precarious situation (perhaps from falling in similar traps in the past), and who end up keeping the card long-term, as the 3-months spending requirement ends up trapping them in the CC spending loop.
This equilibrium you describe, is in many cases built on preying on the weakest and most vulnerable, so no, I don't think it's fine.
You're forgetting about the third class of person, who thinks it's not worth their time, and, frankly, distasteful to spend that much time and effort optimizing, but also finds the offers presented to those that aren't optimizing insultingly poor value.
That's the kind of person that will not download the app to get the coupons and hunt for the good deals at Safeway, but will also not accept the exceptionally high sticker price, and instead will go shop at another store.
That's not true. I'll put numbers on it to illustrate my point, but I hope you have the grace not to quibble with the actual numbers, which are obviously placeholders.
Let's say I'm willing to pay up to $100 for a plane ticket, and value my time at around $50/h.
If the "lazy" sticker price of the plane ticket is $125, but if I put in an hour of work I can find it for $75, the average price of the plane ticket is exactly what I'm willing to pay (and it's likely that the price I'm willing to pay is acceptable and profitable to the company), but they will not sell me a ticket I'm willing to buy.
And that's ignoring any negative value I put on having to jump through hoops to get a reasonable price. It's entirely possible that I'd rather pay $100 for a ticket than spend 10 minutes finding a $75 ticket, even if that's technically worth it for me, because I value my dignity and feeling like the company respects me as a customer. So, I do care, but I'm also negatively motivated by feeling like the company is fucking with me.
> No one with an ounce of respect for their body shops at Safeway regardless of the price.
This is an obnoxiously irrelevant way to dismiss my point, and is why I took this long to respond to you. Please don't do this.
That was actually great. While a Google search is free and easy, that cost real tokens which aren't free (especially for those who don't have a work-paid sub).
Except most people aren't mass spammers, and most people aren't actually heard most of the time, and most of our systems are actually set up to make sure this continues to be the case.
How about AI will show us where we are intentionally underfunding things we have promised people?
If your system collapses under adversarial attacks it is not a resilient system. Bad actors exist and identifying them can be very difficult, especially when you can't instantly violate their rights in many cases.
Amusingly, I totally bought Uranium off of Amazon ~7 years ago when I wanted to test my Geiger counter. Doesn't seem to be available anymore though? Sad!
I'm blind. Apple has explicitly said they wouldn't consider the "bloody ROI" of accessibility (will fund it no matter what). This is important to me. Software freedom is also important to me. What am I to do?
Pick your priorities. There will never exist enough products to check all of the boxes that all of the people say is important to them. There is someone out there who would much rather have a product cost some amount less at the cost of accessibility features but they think a device that 'just works' is important so they buy the iPhone anyway.
The point is that your conclusion here is both the correct/obvious one, and that it ensures one explicitly cannot vote with their wallet. They have priorities too significant to be able to take part in the protest. They are invisible to the company they WANT to vote for.
Perhaps what allows Apple is to be better at certain things like that or security for average consumer is indeed at odds with some aspirations? At least in practice, not utopian dreams, that seems to be the case.
So you vote with your wallet for accessibility despite freedom and the rest of hacker news do the opposite and it all works itself out.
You grab an Android phone anyways because Android accessibility is very much on par or even better than whatever slop Apple is putting out with VoiceOver these days, and developers also somewhat care about it more on Android.
You're also going to benefit from European mandates to have accessible mobile apps anyways, it's coming for free to you.
Even this long is too long. Apple devices don’t just provide better accessibility, they make lives easier by providing better accessibility, Android lagging behind is not just a matter of feature-parity, it’s an entire ecosystem making it harder for people to engage with itself.
- Apple's own UI is widely non accessible and noncompliant. Their toggles are an accessibility nightmare, for example.
- VoiceOver is notoriously non-customizable. It's good enough for blind users, and catastrophic for any motor issues. All of Apple's accessibility is entirely focused over on blind users. Your grandma with the shakes can go fuck herself however.
- VoiceOver is catastrophically unaware of context. I can't count the times it read "Walk fifteen em" instead of Walk 15 meters.
- Apple's solution to accessibility is to not care about their apps and to just make an "accessible" copy.
- Implementing accessibility properly in UIKit/SwiftUI is a kludgey piece of shit where it's extremely common to see people entirely modifying how text is being read, which in an accessibility nightmare and only considers basically blind users.
Source: the software I work on has to be 80%+ accessible, with regular accessibility audits and a panel of various users.
> You're also going to benefit from European mandates to have accessible mobile apps anyways, it's coming for free to you.
Not true, the EU tried to become a regulatory superpower and largely failed to export its rules to the rest of the world. Instead, Apple and Google just aren't bringing new features to the EU anymore.
That was the UK because it was Apple complying with a specific UK law that says the government must be able to access everyone's phones at all times.
The more common one is GDPR. Things like AI features come to Europe late, or don't come at all, because it's illegal to just record everyone all the time and transmit that information back to the mothership, which is what the feature does.
If you know of any company that would bother to spend days making accessibility improvements, yet for some reason decide to not bring those to the US release, they're the dumbest fucking company in existence.
The EU accessibility mandates will be followed by apps, not by platforms. I don't care if Apple Music wants to keep being trash, Spotify, Deezer and Tidal will improve and lift everyone with it.
I wasn't convinced by that explanation. Yes Cornell claimed that Elias the Lighthouse Keeper came from their WildChat dataset, which is a million conversations with GPT 3.5. But there are some problems here:
• They claimed this in a press release, not the actual paper they published. Suggests lack of confidence in the claim.
• Elias only appeared a hundred odd times in a million conversations. They have no viable explanation for why such a tiny amount of data would cause a totally skewed distribution of outputs, other than "it's a safe story" and "trying to avoid copyrighted characters", which is hardly an explanation at all?
• It's Cornell claiming that Cornell's work was important. Obvious risk of motivated reasoning.
• It implicitly assumes every major lab is distilling from ChatGPT without saying so. That's a big leap, given how much OpenAI and Anthropic hate distillation. Is it really plausible that Anthropic and Google are effectively distilling ChatGPT without realizing/admitting it?
• It seems to invert causality. WildChat came from an LLM, so if there's something latent in LLMs that makes Elias appear, you'd expect to see it in that dataset. Why do they assume that dataset caused the problem rather than merely show it? The obvious answer would be because the stories were told to the LLM by humans, but they don't claim that. They only say the character appeared in the dataset.
Hum, I'm actually pretty sure that Text-davinci-003 used to talk about Elias all the time now that you mention it. I was doing stuff with NPC generation around Dec 22 and he was every third character.
Hmm. If so then it suggests Elias has been an obsession from the first LLMs ever made. And text-davinci-003 can't have been trained on a dataset of LLM conversations either.
I would have been tempted to argue something like this: maybe OpenAI accidentally overfitted RLHF such that, like, one Kenyan contractor wrote a story they invented and they just happened to pick Elias and Elara, then the intermediate DNN fixated on that because of a lack of RLHF samples for fiction and that propagated through every model trained since because they don't randomly initialize the weights, but start each new pretraining run with the old weights adapted. So one guy doing RLHF work in the early days somehow propagated this story.
But that can't explain why every lab ended up in the same place, why models all converge on the same ideas - even image models. So... back to the drawing board. There's something weird about high dimensional space that is unexplained.
I do not look forward to when rules like this run up against a11y concerns. The AI equivalent of the Domino's case will certainly be interesting!
reply