I looked into upscaling for the original episodes of Police Squad. Paid a good chunk of money for an AI upscaler because I was too curious how good of a job it could do.
The results were not great. When there's too many different things impacting the quality of the source, the AI just can't do a great job. In the case of Police Squad DVD rips, they were low res and noisy.
When I eventually found a higher resolution Bluray version, I ran that through just to denoise it and got much better results. My understanding is that it also would have been able to do a good job upscaling given a source that wasn't noisy.
Maybe you can use a chain of models to handle both problems. I didn't try anything like that. Though I would expect the tool to have that as an option if it is something that worked well.
Repairability was bad but has gone up a lot, they did a big internals redesign for iPhone 15 to make it easier to swap batteries and replace screens. Still not something they want user serviceable, I imagine mostly because it creates headaches for everyone involved. Most people struggle fixing big things, let alone sub-mm precision things. But this helps the 3rd party repair shops a lot.
And at the time they were comparable to a flashlight in terms of complexity, not running AAA games with raytracing and a camera pipeline of untold complexity. It’s almost like having anything this complex working requires insane engineering and miniaturization, this is not due to “planned obsolescence”, especially if you take a look at the second hand market. No other brand has even remotely similar resell value.
Yep they do this for all their platforms, set a battery target based on ~90%tile daily usage of particular form factor (e.g. 18hr watch/iphone, 10-12hr macbook/ipad) and then let ID & chip teams do whatever they want to hit that. The value add of every additional minute/hour of battery life drops off from there because daily loops for charging are realistically what everyone will do, and there's more value add in other places like making things thinner, faster, or do something new.
The argument about LLMs is wrong, not because of reasons stated but because semantic meaning shouldn't solely be defined by the publisher.
The real question is whether the average publisher is better than an LLM at accurately classifying their content. My guess is, when it comes to categorization and summarization, an LLM is going to handily win. An easy test is: are publishers experts on topics they talk about? The truth of the internet is no, they're not usually.
The entire world of SEO hacks, blogspam, etc exists because publishers were the only source of truth that the search engine used to determine meaning and quality, which has created all the sorts of misaligned incentives that we've lived with for the past 25 years. At best there are some things publishers can provide as guidance for an LLM, social card, etc, but it can't be the only truth of the content.
Perhaps we will only really reach the promise of 'the semantic web' when we've adequately overcome the principal-agent problem of who gets to define the meaning of things on the web. My sense is that requires classifiers that are controlled by users.
Yet LLMS fail to make these simple but sometimes meaningful differentiation. See for example this case in which a court reporter is described as being all the things he reported about by Copilot: a child molester, a psychatric escapee, a widow cheat. Presumably because his name was in a lot of articles about said things and LLMS simply associate his name with the crimes without making the connection that he could in fact be simply the messenger and not the criminal. If LLMS had the semantic understanding that the name on top/bottom of a news article is the author, it would not have made that mistake.
Absolutely! Today's LLMs can sometimes(/often?) enormously suck and should not be relied upon for critical information. There's a long way to go to make them better, and I'm happy that a lot of people are working on that. Finding meaning in a sea of information is a highly imperfect enterprise regardless of the tech we use.
My point though was that the core problem we should be trying to solve is overcoming the fundamental misalignment of incentives between publisher and reader, not whether we can put a better schema together that we hope people adopt intelligently & non-adversarially, because we know that won't happen in practice. I liked what the author wrote but they also didn't really consider this perspective and as such I think they haven't hit upon a fundamental understanding of the problem.
Really depends on what sort of person you are I guess.
Some people appreciate being shown fascinating aspects of human nature. Some people don't, and I wonder why they're on a forum dedicated to curiosity and discussion. And then, some people get weirdly aggressive if they're shown something that doesn't quite fit in their worldview. This topic in particular seems to draw those out, and it's fascinating to me.
Myself, I thought it was great to learn about spontaneous trait association, because it explains so much weird human behavior. The fact that LLMs do something so similar is, at the very least, an interesting parallel.
>My guess is, when it comes to categorization and summarization, an LLM is going to handily win. An easy test is: are publishers experts on topics they talk about? The truth of the internet is no, they're not usually.
LLMs are not experts either. Furthermore, from what I gather, LLMs are trained on:
A different name offers a different perspective, because of all the associations with the name. Problems that are hard to solve are often hard because we're stuck on a particular perspective as to how to solve them. Reframing with new associations is a way to gain a new perspective, to look at the problem differently, to gain insight that you previously did not have. This is an extremely common and effective problem solving technique.
just bad unreproduceable psychology research. there is zero proof of this. we actually have examples going the other way though.
the idea that changing the words used can change your ideas about something is a weak form of the sapir whorf hypothesis, which is close to universally panned in stronger forms and highly suspect in weaker forms except in pop psychology.
pinker called a similar idea - changing a word to avoid previous negative connotations - the euphemism treadmill: from retarded to handicapped to disabled to differetly abled, but never changed anybodies views. they just carried them over. its because language is a reflection of our inner thoughts, not the other way around.
That's weird because I find value in using this technique in both personal and creative contexts, and these kinds of reframes are used all the time in therapy, in school, with parents talking to their kids, etc.
Perhaps the finer points of those studies are not applicable to the topic at hand, which is an individuals strategy to gain new perspective on their own problems, rather than the nth-order effects of proxy words in culture, or researcher's anthropological interpretations and comparisons of languages effect on worldviews across extremely different cultures.
I get that it's a popular topic on HN to point out the replicability crisis in psych research, but the nature of the beast of a high level / subjective / messy subject like human psychology is that you have to be extremely precise about what you're testing and what conclusions you draw, or you're at risk of generalizing beyond the data. What you've cited has surface level similarity to the topic at hand, but is quite different in the specifics.
Plus it doesn't even stand up to the sniff test - language impacts us. Words impact us. The subtleties of how language is used can have profound effects on how we live our lives. Haven't you ever read a beautiful sentence over and over, or marinated in an obscure word and all its intricacies? This is a common sense proposition.
I second this. I could have said I "agreed" with this, but to "second" this conveys something different. Or rather, from my intent it seems to convey X but maybe you receive it as Y.
The challenge I find with so much language is the vast number of associations we carry with words and how connotations can vary so extremely even amongst people who "have the same background."
One of the best descriptions I heard for language I think was written by James Pennebaker talking about expressive writing and how words were basically putting a digital categorization onto an analog signal of experience.
Words are not very precise and are often very relative approximations that require so much negotiation to reach shared meaning. Some will read "quests," as I did, and immediately think of Monty Python and the Holy Grail and feel a bit goofy and have a hard time saying "I'm going on a quest" seriously. Others will feel excited and maybe encouraged to see it that way. Others might be annoyed because they love the word "goals" and have built their brands and careers around the word "goals." At the same time, with repeated usage of the word "quests," even my emotional reactions to it may change and I start to embrace the word with more seriousness.
---
> just bad unreproduceable psychology research. there is zero proof of this. we actually have examples going the other way though.
For example, I read something like this and in my head, I often reframe it. I read "just bad unreproduceable psychology research" as "I do not trust the research you are quoting because I think it is not reproduceable." I read "there is zero proof of this" as "I have not seen any proof of this or do not believe any proof exists." I read "we actually have examples going the other way though" as "I have seen examples that seem to dispute what you are saying."
The way it was originally written, in terms of word choice, seemed to describe to me an objective truth in the universe, whereas the reframe I applied shows more of a relative belief that you may have. That's not to say your beliefs are not the capital T truth, but rather for me to feel less angry when someone tells me "how the world is" and to try to see the world from their perspective and learn from it.
Hey Jim! Yeah I think you hit the nail on the head, the difficulty of measuring language's impact on us comes from the individuality of response to it. Even minor shift in situational context can alter our response, as can the measurement window (e.g. you can see yourself warming to 'quests' over time), and of course the actual precise stimulus (our response to close- vs open-ended framing, process vs outcome framing, and the individual's cultural knowledge of the specific language used). Thus it's very hard to generalize these snapshot-in-time personal experiences across populations.
A recent comparison might be SSRI's effects, which are proving to be no better than placebo at a population level, yet a large body of individual anecdotes show they have very positive effects on some. Rather than dismiss the anecdotes, we need to acknowledge the difficulty of measurement for such a complex & high level topic and be curious enough to look at the problem through different mechanisms of ascertaining truth like qual analysis, logic, and common sense, rather than just accept some murky methods and results as the final word.
> A different name offers a different perspective, because of all the associations with the name.
The associations are often misleading.
For example, when someone says "sprint", my pre-IT associations would be "run as fast as possible for a very short time, then take a long break", but everyone knows that this is not how agile development works.
(Ironically, the IT word for "work hard for a relatively short time, then take a long break" is "hackathon".)
For an example of how these may be used, Kialo [1] uses a form of argument maps for structured debate. There's also an Obsidian plugin for argument maps [2], tho it's a bit out of date.
I love the tech but these things are going to be obnoxious as hell. I hope it supercharges the movement to ban advertisements from public places, which has already been done successfully in Sao Paulo and several other major cities [1]. The 99% Invisible episode on it is worth a listen [2].
Many years ago São Paulo heavily regulated outdoor ads [1]. It was a huge leap in quality of life for the citizens - the removal of all that visual noise and the sudden reappearance of the underlying architecture unveiled a beauty most inhabitants didn't remember existed.
It got to the point that, when I visited New York, I got angry at the full-on sensorial assault. My mood was always nasty there.
If this is done responsibly, and used to extend architecture in ways it wouldn't be feasible before, I don't have many reasons to be concerned. OTOH, in places where visual pollution is unregulated, oh boy, it'll be bad.
Probably, but we'll probably get used to it just like graphics on windows or displays behind windows / store front displays. Sci-fi and cyberpunk visualisations have already prepared us.
First class - no ads. Second class, cheaper & ads. If it goes this way its actually great in 1 aspect - you can't escape the simple fact that ads are a garbage, and high quality of... life or your time is without them.
Seems obvious but I've met many who dont see it as sort of brain cancer and consider them harmless.
My last visit to NYC there was a pronounced uptick in wall sized LED screens being used for everything from art pieces (at the new Laguardia) to info screens to just shops with massive screens in their storefront in lieu of showing physical goods.
Talk about a profound misunderstanding of the powers that shape the tech landscape and open consumer wallets.
Bundling, convenience, tight integration, and bringing technologies to larger markets are like the most powerful business moves you can make in tech. Making something one tap away instead of four unlocks at least an order of magnitude more engagement and usage.
I think you are wrong and the lackluster sales of AI-enabled iPhones will prove it. The exact same thing happened with Siri, and was she a system-seller? No, people actually switched phones because Siri was the only assistant you could use. Mandating a one-size-fits-all solution that everyone hates does not automatically make your solution good again. This exact pattern is destined to happen with ChatGPT; the people that want it already use it, and the people that don't care won't be happy about being forced to use a solution that doesn't work for them.
Siri was a system seller - The iPhone 4S that Siri was introduced alongside sold 4mm units on the first weekend which was the fastest selling iPhone to date. This was an 'S' update where the body style had not been changed, so the draw of the device was purely related to internals updates and Siri, which was heavily promoted. I don't know how long that sales bump lasted for but I don't think you're correct that Siri wasn't a system seller.
Few people switched because of Siri - I'm not sure in practice whether any significant amount of people switched away from iPhone due to Siri, that sounds like something that might be true in niche circles but probably not something that made a dent on the general market of >1 billion people and devices. Even if a million people switch thats a rounding error, and frankly HN is famous for overestimating the techie-ness of the general market.
ChatGPT isn't a substitute - ChatGPT is just one way AI tech is articulated. Most people don't like chatbots (see late 2010s hype like Messenger bots that died fast). Most people don't have time to tinker with AI tools to know whether it will help them in their lives and all their little use cases. Integrating AI into existing workflows on iOS will expose it to an order of magnitude more people. If they like the particular use cases it enables, maybe that'll move units. It will require a lot of user education in the form of marketing to get people to see these capabilities, because again the general market isn't tinkering with AI and doesn't know what it can do. Regardless, your point that any who were interested would use ChatGPT makes little sense because ChatGPT isn't a substitute in practice for general market users, nor the specific use cases Apple showed off where they use AI, due to tight integration, different UX, greater convenience, etc.
Totally agree. You can't reduce an experience like Minecraft to any one thing, it's the formula that makes it so enduring.
re: harping on Minecrafts graphics, there's a lot of people that completely misunderstand the purpose of graphics in games and just think greater fidelity = better. Sure, in AAA games especially, graphics are there to sell the game, and that's certainly one part of the creative ensemble that the medium is known for.
However, graphics also enable game designs. These include creating a sense of grandeur or plausibility (high fidelity), making fast feedback loops easier to see (high contrast effects), allowing for visual challenges (hidden items), freeing resources (low fidelity), and probably most importantly, drawing attention to relevant parts of play.
The best creative decisions are those that tend to check a lot of boxes with a single decision, and Minecraft's lo-fi style really did that. It enabled complexity within performance constraints, drew attention to the resource game and away from the visuals, defined it's own aesthetic reminiscent of older games, and allowed the game to be built with a very small team without a lot of art skill. Basically notch took lemons and made lemonade, and it worked.
My favorite part of the Minecraft aesthetic though is that it implies you can create, making the whole world creatively legible with big blocks. Other creation games in 2d & 3d just don't have this degree of creative suggestiveness, they look too much like other read-only games to make you feel like you're an equal participant in play.
Minecraft is the best building game ever made. It's perfect like Bach - built so simply you can see how it was done, and fits into the constraints of the medium perfectly.
A game that has similar building (albeit less "scriptable" since there's no redstone) is Enshrouded. They use smaller voxels and a different polygonisation algorithm (seems to also differ based on voxel material tool). It's a really fun game, even if there's less variety than Minecraft. Also it's still in early access. I recommend you check it out if you enjoy building in Minecraft
There can also be too many graphics and effects going on. I played the Elden Ring dlc and a few fights there are so many particle effects going I literally have no idea what's going on sometimes.
reply