tavavex

> Problem is, because statistics, 90% confidence is wrong 10% of the time. Now you have a big polarizing news story about Apple pushing Trump (or Biden) propaganda onto some maximally sympathetic person who hates that sort of thing.

But think about it, we're not talking about auto-suggest here. The distinction I made in the parent is critical. Can you imagine what the article headline is like? "I swiped 'duck' into my iPhone. It said 'fuck'." - something like that? It's ludicrous at face value, especially because you can't show this was on Apple in any way. The immediate reaction from the readers, Apple and any random bystander would be "lol the author must have swiped it in really weird". I just can't see any variation of this, even on the most divisive topics, doing any actual damage to Apple. Unlike autocorrect, it just comes off as your phone 'mishearing' you and outputting something similar that just happens to be problematic in some way, not as Apple intently putting words in your mouth. If we assume the best of intentions from Apple, they're still overreacting and purposefully setting arbitrary hard rules about how you can't use their (not your) device.

I'll restate a lot of what I have said in this thread that I think is still very relevant to what you said.

1. I realize that swipe input is probabilistic as opposed to using a physical device. I just think it should not matter - if it's probabilistic, they should have specific hard rules that are obeyed in all cases. For example, if the software concludes that the input is likely to be this dictionary word with 95% certainty, it should output that word. The context surrounding the word can be used to fine-tune the certainty, but not to cross-check against blacklists of banned phrases.

2. The reason for #1 is that if the CYA approach is allowed, there are no arguments left to prevent Apple from looking at your entire paragraphs of text and then dynamically deciding to discourage certain predictions, to prevent their device from being used to create something they don't agree with.

3. Additionally, because typing on a digital keyboard is also predictive (the touch zones of keys are weighted depending on what it thinks you're likely to type next), this would also allow them to literally make typing certain 'unsafe' things more difficult.

To me, there is a big difference between an Apple algorithm saying "we saw what you typed, would you like to follow that up with <suggestion>?" and "we know what you wanted to type with 90% confidence, but we wouldn't want to say things like that, would we now?". One is predictive, the other acts after the fact.

If Apple doesn't want its autocorrect to suggest statistically likely opinions on divisive topics, then sure, whatever. But when it has a hard blacklist of input it will categorically refuse from the user and editorialize it with nonsensical changes that only serve to avoid the blacklist, that is a very, very bad look.

I just don't think you understand what soft coercion is.

You have door 1 and 2 presented to you by me. You're equally likely to want to pick both options. Door 1 is open. Door 2 is locked, but you can ask me to come over and undo all the locks to open it for you.

Run the test over a large number of people, and then shockingly, it turns out that doing this made more people choose door 1, even though they didn't have a preference beforehand. According to your logic, I did not coerce anyone by designing the choice in this obtuse way because after all you can still choose between the two options. But more people ended up choosing the option I wanted them to pick, weird.

If this coercion didn't work, companies wouldn't be hard at work designing revolutionary new types of dark patterns that make you more likely to accept the cookies, back out of deleting your account or decline enabling an option in the settings they don't want you to use anymore.

I would be shocked if an LLM is involved in any step of the process here. I guess the odds aren't zero, but swipe input is well over a decade old at this point. I see it as essentially a solved problem that's entirely doable with standard algorithms - not models, let alone the large language kind. Maybe some model could be used to analyze the context and shift the weight slightly to more probable phrases, but at the core, when you look at inputting single words, it's just checking the letters you swiped over, input timing, turning points and adjacent letters against its dictionary.

A lot of input methods are probabilistic in their nature. Some other examples are voice to text and even regular digital keyboard input (fat-finger correction that tries to pick the letter you wanted to tap instead of the letter that you actually tapped). So even with your distinction, actually spelling it out with the digital keyboard would not make a difference because the software still has to make a call there. Even if you plugged in a physical keyboard, Apple could still say that the signal passes through its software, which can constitute endorsement, so it should run the checks against that too, just in case. Nothing in your argument stops that.

I fully reject that this makes the situation any better or more justifiable. The job of the software that picks what you have likely just said, swiped or tapped should only concern the given word. If the algorithm decides with 95% certainty that the input was "dictator", that's what it should output. If context is used to better select the candidate word, it should only be done for the purpose of slightly re-weighting the possible options with one transparent algorithm, not for checks against a blacklist.

The reason why I think this is that your line of thinking is unbounded, and there is nothing stopping the next logical move - if Apple can decide 'not to endorse' (representation isn't endorsement, but whatever) views by blacklisting its algorithms from forming certain phrases, why can't they uplink entire paragraphs of your text as you're typing it to dynamically analyze them instead of using a dumb filter, and then decide if the entire text is something they want to allow?

But what I'm saying is that they're not going to become 2-5x cheaper if the constraining factor isn't supply. The companies are definitely not acting they're short on compute - otherwise you'd see queues, much more stringent rate-limiting, paying for priority and similar measures. I maintain that the buildout is more about servicing hype and crazy projections than what we actually need today.

I wonder what the boardroom meeting for this was like. Despite the outcome, I doubt the devs were pitching censorship as the reason for implementing this.

To the shocking number of people here who are justifying this or explaining how it only affect swipe input as if that makes it better: soft coercion is still coercion. If I made a physical keyboard that stops keys from working when you're about to finish typing something I don't want you to type, but you can still circumvent it, I still made a censorship keyboard. It's a sign of the times with how used tech companies are to 'softly' moderating societal trends by unilaterally tweaking the software that everyone uses.

> Not building out data centers would mean your $200 spend would be $500 no? Just supply/demand wise

Is there actually any evidence we're short on inference compute, let alone by a factor of 2.5? To me it always seemed the data centers were about accommodating some extreme predictions on future demand, scaling up training and trying to outdo one another in capturing the limited hardware.

Usually, people tend to care about what the overwhelming majority thinks, if only to remain informed. I guess branding all dissidents with a derogatory term of your choice and retreating back into the booster bubble is an alternate option.

> Tech is now viewed in the same lens as dirty greedy oil companies. If an oil company opened a candy store most people I know would not go because "Oh those oil companies can't be trusted"

That's right, but consider what people would be saying about oil companies on Driller News. Thousands of users here have financial or ideological vested interests in the success of the current crop of tech companies. They can recognize the individual faults of the industry where it doesn't affect their primary interest - such as disliking social media, addictive design and so on - but when you start talking about THE tech golden goose, the criticism dies down.

Thinking that people should decide if an infrastructure megaproject of dubious utility to their community should be built? Basically identical to thinking that people should decide what races live in their community and whether slavery is allowed.

The HN crashout over people suddenly disliking hyperscalers for some reason is something you could only witness here.

I'm one of those younger people. Having graduated from a CS program recently, I think that the main point of that degree was to teach me a wide range of relevant concepts and ideas in math and computer science that should allow me to reason about lots of different topics in the field.

I don't think the degree should be optimized toward hireability, but to me its value has still greatly diminished. My assumption was that the CS degree wouldn't follow every whim of the industry, but will still give me enough skills and knowledge to be able to pick up anything CS-related without too much extra training. There's no promise of a job, but there was an expectation that you were learning something useful and versatile that you could then turn into a tech job with some added work. Now, the value of all that has mostly vanished.

Instead of just one tech niche being automated away, leaving the others relatively intact, AI has plugged every opening that people like me could get into at the start of their careers, and this effect is the same across almost the entire industry. And for the people who got in, their employers are already valuing those fundamentals less and less as workers are encouraged to paper over any gaps in knowledge with AI. The relevance of all those learned skills has fallen. I think the only people this doesn't negatively impact are ones who are dead set on going into academia and becoming a researcher in a CS field. And maybe not even then, as AI seems to be taking the spotlight from most other research and is getting shoehorned into every new idea. Thinking the degree is a job training exercise and expecting your core skills to be relevant after graduation at all outside academia are different positions.

Can I ask about the uniform width formatting? Is this some convention from the old days, like formatting text to perfectly fit on a terminal screen?

Whenever something happens that you don't like, make sure to set a timer for a year so you can be conveniently reminded at precisely the moment when you must flip your worldview and fall in love with that thing or event.

Is there any evidence that we're short on compute for the AI demand? My impression was that many data centers were being planned to serve hypothetical future demand (with the assumptions that it'll multiply many more times while also not becoming much more efficient) and to facilitate model training, neither of which has much to do with the argument you're making.

According to your basic economic theory, none of us should have ever gotten any inefficient personal goods if they can be pooled into more efficient clusters. We wouldn't get personal computers, because obviously renting time from a large computer is much more efficient. We wouldn't get personal transportation, or non-communal housing, or inefficient small businesses. Are you sure you want to argue this point with all its logical consequences?

Let me know when you're ready to move onto advanced economic theory, including what happens to efficiency and value when a tiny number of entities concentrate an in-demand commodity in their hands and refuse to share.

Damned if you, damned if you don't. Update constantly and you might get a fun little surprise shipped to you in one of the updates. Update occasionally and you'll miss the critical fix(es) to a vulnerability you have.

I think the only way out is extreme scrutiny of all updates and paranoid minimization of all attack vectors, but the industry is only going to settle into that local minimum after we go through some devastating breaches.

> Yes, you can create garbage with AI tools. You can create garbage using Blender, Unreal Engine, Unity, GIMP, Ableton, iMovie and Word

Do you not see any difference between creating garbage in these ways?

People have created low-quality things since the dawn of humanity, using all sorts of tools, professional and not. However, before AI, making most types of garbage still required some human work, it wasn't infinitely scalable. And what was automatable ended up being so horrible that most people could easily dismiss it on the spot, or for them to even be filtered out algorithmically. On the internet, it wasn't perfect, but the signal-to-noise ratio was still at a level where you reliably discern the signal part.

AI changed this. Even a beginner can set one up to run almost fully unattended - and while what comes out won't be good, it also won't be immediately apparent as such, especially to the general public. More and more I find myself being fooled by a website link, a YouTube video thumbnail or a paragraph in an article. AI lets people create garbage that can impersonate real effort decently well, and they've used that to great extents. As we speak, thousands of agents are working day and night to pump every online platform with their 'content', without a single person in the world to care about what they put out. They're taking over and removing any good signal we had, everything becomes diluted with trash on a scale we have never seen. People notice that. Many are clueless about AI, and social influence plays a role, but they're not being misled and have their opinions manipulated. They wouldn't enjoy most of this content regardless.

In addition to the dilution of good information, they care about appearances and how AI lets you create an appearance of caring about something. The cheap food stand seemed honest when they put up an unassuming print of a simple Word document, and now seems dishonest when they put up an AI-generated mess that looks nothing like the real product and desperately tries to seem more refined than they are. Their manager was at least honest when he replied with a lazy 'k', and now he's lying about caring by dumping out a bullet point list of AI-generated text for a response that we know he himself hasn't read. It's the dream product for scammers, grifters and people who simply don't care.

Can you use AI to help create something good? Sure, but anyone doing that still needs to take their time and put care into it. Even if AI makes them faster, by the time they finish, 10TB of more slop has already been produced to drown them out, because that can be fully automated. Effectively 99.9% of AI content ends up being garbage, so of course people are going to reject it. Most probably don't want the few nuggets of gold for the ocean of waste.

You know what else will lead to us not getting your cures or space travel? Fully automating enough people such that there are no longer enough jobs in society to do for the majority. That would create a destitute underclass that can never get employed and earn anything to survive, and a tiny elite class that can truly live and afford a fully automated lifestyle, space travel, etc. We are well on track to making this happen within this century. The only way to avoid this is to share some of the value of removing people's jobs with the people it displaced. People doing delivery work are already at jobs of last resort, how do you think "sucks to suck" fares for them as an argument towards removing their livelihood?

Ludditism is not 'a disease', how dehumanizing. It's an expected response. If we did anything towards sharing the profits, it would not exist. If we answered the hard question of "how will a person survive if we permanently cut the number of all jobs in half?", it would not exist. Ludditism for the people being made irrelevant is stepping out of the oncoming train instead of proudly letting yourself be run over to take one for the team so maybe in a few decades' time some of the added value of new automation cancels out the job loss (how it was in the past - and that was when automated jobs still required loads of brand new workers).

> So what's their selling point then?

Same one as almost every other modern tech company. AI AI AI AI AI AI AI AI AI AI AI

> company-wide ‘re-interviews’ that would have forced employees to explain how they are using AI, with the clear intention that they may be fired if not using it enough in the eyes of Proton leadership

This is evil in ways most people probably don't understand. To explain, here is the real message behind this, stripped of neutral journalism language. I'm not going to be shocked if this is the actual hidden thought process.

"Hello. You have now worked at our company for 7 years, contributing thousands of hours of work towards making our products better. We regret to inform you that you are worthless. Your so-called 'contributions' don't deserve to be printed on toilet paper our CEO wipes with. Frankly, you are a waste of space sucking up the air inside our offices, and we are itching to get rid of degenerates like you at the first opportunity. Your only chance to temporarily offset our perception of your uselessness is our upcoming 20-minute degradation ritual where you'll dance to our whims and plead by showing how much you love using whatever our CEO says you'll use. Building good solutions, knowing our tech stack, working with other people, being creative? All worthless, just like you. What really matters is if you can make me feel good. Fuck you and get ready to crawl on your knees."

No matter what your opinion on determinism is, the point I made that what OP is doing is very unlike a predictable and deterministic factory game still stands.

But beside that, I don't think that LLM non-determinism is really the same thing as what you see in engineering. Engineering has things like tolerances and redundancy, and while they accommodate random events, they're known, estimated and expected. They also deal with external factors, not the potential of shoddy engineering in the same system they're working on. The best way I can compare it is building a bridge that can withstand twice the normal expected full load and with enough structural redundancy to not collapse even after a prolonged lack of maintenance. These things account for many types of natural degradation and edge cases that are expected. And then there's building a bridge that you know will collapse in the next 10 years at a completely random moment, and instead of engineering a more robust solution, tacking on more gadgets to make the collapse less bad - automatic gates that close the entrances to the bridge, lake drainage system that will suck out the debris, a drone system that drops a brand new randomized bridge on the same spot, and so on. It doesn't feel very much like engineering.

If that doesn't sound convincing to you, consider one other thing. If you gave an engineer a non-deterministic solution and a deterministic solution to the exact same problem, which do you think they would prefer? Would they willingly remove determinism from a system? Maybe it's not the chaos of the real world that scares software developers so much - lots of them have to work on systems that directly interact with the real world. What if it's the fact that there is now a fight to take away their nice, understandable, satisfying deterministic solutions and inject probabilistic chaos into things that never had to have it in the first place - into absolutely everything?

Unless you watch lower-quality content that tries to heavily pad for time or contains a lot of unnecessary content (such as tutorials where only 1 minute in the video matters and contains what you're actually looking for), summarizing 20 minutes of information into 5 bullet points feels like watching a movie or reading a book by skimming the synopsis section on Wikipedia.

Not really, they're not ripoffs. Maybe loosely inspired, but besides the hexes and the theme, they look to be mechanically different. The board game is a competitive versus game, not a set of puzzles; it has varying numbers of voters per hex; the game board is of a preset shape and not freeform.

The game in the post is basically identical to Gerrymandle, they're far more similar to one another than either game is to what you've posted.

Really? So I can ditch Minefraft for Infiniminer and get an identical experience? Same mechanics, items, bosses, worlds, achievements, structures and all?

Or are you perhaps using a game that was merely inspired by another concept but expanded on it in brand new ways to defend a game that is literally identical to what it's copying?

You can get a feel of the state of the software industry through the replies to this comment. Specifically, the countless times people counter with "Oh, well everything is a copy of everything", citing some X that's just like Y. Except X is merely inspired by Y and in reality has far more original ideas that make it completely different. In this case this is a literal carbon copy, and people are automatically rushing to defend it. When the 'vibecoders copying indie games 1-for-1 and reselling them on Steam' debacle happened, these people probably wanted to congratulate the ripoff makers for their good business sense and brilliant use of AI.

It's 66% now. But that only makes my point stronger - why is a shovel-selling company that has been doing so well (at least as far as finance is concerned) having a one-day wipeout like this on relatively mundane news?

Seems that most articles agree that if you look at just the fundamentals, the only thing that was bad that Intel reported was that they're promising less profit next quarter than expected. Despite their server division booming, the PC crisis is holding them back a bit. But I don't know if this is a reason why it would drop this much, maybe companies are using this to price in them vastly overestimating Intel earlier this year or fears about the future of its AI division.

I'm no economist or investor, but down 28% in a day seems insane for a company at this scale. And they didn't even perform badly, imagine what would happen if they missed their earnings projections.

No offense, but this feels like the most un-Factorio thing imaginable.

In factory building games, the systems in play can be understood and are fully deterministic. You're given a set of tools and a problem, and then craft together a solution. Your reward isn't just seeing the item you want on a conveyor belt, but looking at the factory you built ticking like clockwork.

What you're describing is beating out endless nondeterministic fires. It's like you were forced to play Factorio exclusively through an algorithm that designs incomprehensible factories that don't work sometimes, but designs them much faster than you ever could. So you click the magic button and see a mass of buildings appear in front of you. They may work perfectly fine, but if they don't, you need to use the same statistical factory estimator to build an in-game factory fixer mechanism and factory tester mechanism. And then a few updates later the factory fixer breaks so you need to deploy a factory fixer fixer, and a belt clean-up machine, and so on ad infinitum. To me this doesn't seem satisfying in the slightest.