Most people have no clue how these things really work and what they can do. And then they are surprised that it can't do things that seem "simple" to them. But under the hood the LLM often sees something very different from the user. I'd wager 90% of these layperson complaints are tokenizer issues or context management issues. Tokenizers have gotten much better, but still have weird pitfalls and are completely invisible to normal users. Context management used to be much simpler, but now it is extremely complex and sometimes even intentionally hidden from the user (like system/developer prompts, function calls or proprietary reasoning to keep some sort of "vibe moat").
> Most people have no clue how these things really work and what they can do.
Primarily because the way these things really work has been buried under a mountain of hype and marketing that uses misleading language to promote what they can hypothetically do.
> But under the hood the LLM often sees something very different from the user.
As a user, I shouldn't need to be aware of what happens under the hood. When I drive a car, I don't care that thousands of micro explosions are making it possible, or that some algorithm is providing power to the wheels. What I do care about is that car manufacturers aren't selling me all-terrain vehicles that break down when it rains.
Unfortunately, cars only do one thing. And even that thing is pretty straightforward. LLMs are far too complex to cram them into any niche. They are general purpose knowledge processing machines. If you don't really know what you know or what you're doing, an LLM might be better at most of your tasks already, but you are not the person who will eventually use it to automate your job away. Executives and L1 support are the ones who believe they can benefit personally from them the most (and they are correct in principle, so the marketing is not off either), but due to their own lack of insight they will be most disappointed.
You can't move them (apart from the opening and closing animation), but they can move other objects that are in their way. Both need to be physics objects for that to work, even though the door is just kinematic (i.e. it won't react to forces applied to it). Although if I remember correctly, they are not even fully kinematic. I think you could get them stuck halfway closed by cramming something in the door frame that would get the whole thing jammed.
> I think you could get them stuck halfway closed by cramming something in the door frame that would get the whole thing jammed.
This was a popular griefing tactic when TF2 first came out where you could trap everyone in spawn by crouch-jumping into the spawn door as Scout: https://youtu.be/JUPzN7tp7bQ?t=243
This. If you want someone with honest insight into high energy theory, you probably shouldn't listen to a person who still pushes out papers that favour MOND over DM by curve-fitting special galaxies. This topic has been studied to death and it simply doesn't work well enough to warrant such strong opinions against DM. That's why her papers mostly go ignored in the community (many of them aren't even physics anymore and edge more into philosophy-babble or even sociology). All they do is make captivating conspiracy youtube videos for the average layperson.
There's a reason why worker owned companies (which were quite prevalent and a key factor in the 20th century's socialist movements) did not survive in our capitalist world. If we want to see these kinds of companies thrive, we need to change the entire market philosophy. Otherwise they will always be out-competed by companies that favour revenue/profit over worker benefits. In fact they will not even take off in the first place, because that needs capital and investors who own capital will want shares in return, which goes against the core principle of worker-ownership.
I am particularly interested in coops as a model for a tech startup. How would one go about structuring a new corporation to best ensure it remains a coop?
Plenty of worker owned software consultancies in germany, and specially in Hamburg for some reason.
From the site of one that I that I used to work for(they are very friendly so you can hit them up if you want some advice on setting up one)
> How exactly are you structured?
> In our search for a structure that consistently implements the principles of responsible ownership, we came across the veto-share model. First, the principles are enshrined in the company's articles of association. Then, company shares are transferred to a controlling shareholder. This controlling shareholder is granted veto rights, which must be used to prevent any future deviation from the principles. We are delighted to have the Purpose Foundation on board as our controlling shareholder!
> We want to offer every (new) team member the long-term prospect of assuming entrepreneurial responsibility as a co-owner. The criteria for this are already defined in the articles of association. To simplify joining and leaving, we have established dyve Trust eGbR, a partnership that holds 99% of the voting rights in dyve.
Mondragon is often used by people who don't know much about the subject as a counter example, but if you look under the hood, you'll see it doesn't even compare: Less than half the workers own shares and the majority of shares is held by a small circle of executives in the richest branches. That's basically a standard capitalist pig with some socialist lipstick to give pseudo-Marxists an argument in these discussions, when there is none to be had in the first place. In fact Nvidia has a higher percentage of employees owning shares than Mondragon. Would you call that "worker-owned" too?
None of this should be surprising either, because anything that truly benefits workers will be at odds with things that benefit the business. In society there are always people who primarily get paid and people who primarily work. Even the real Marx realized that. If you really want to defeat capitalism, then you have to do more than just play pretend.
There are many successful co-ops all over the world, including in the US. You just don't hear about them because they don't look much different from corporations from the perspective of the average person.
The main they don't survive is not due to capitalism. (Capitalism is just trade without use of force at the end of the day) but due to the lobbying power of corporations. Corporations constantly lobby the government for more regulation and rules, which they just budget for, but smaller companies don't have the funds to meet so they cannot stay in business.
Look at the first couple of years of COVID for example, Walmart, Target and the big box stores lobbied the government to stay open because they were "essential" while all the small businesses got wiped out due to their lack of lobbying power.
The solution is to stop giving corporations the ability to lobby governments, and stop using the government to control/fund markets. Not to blame it on capitalism.
> the small businesses got wiped out due to their lack of lobbying power
Small businesses have a lot if lobbying power (NFIB, US Chamber of Commerce to a certain extent) especially during the pandemic. The PPP loan forgiveness program was ridiculously generous. Despite it being marketed as a way to save jobs, only a third of the money ended up going towards jobs that were at risk.
Those are good points, but it again shows how government getting involved does more harm than good.
The government forced the lockdowns instead of letting people make their decisions about their health, then they tries to aid the businesses they caused to suffer, and most of the money didn't end up where it was supposed to. So many examples of this over and over when you look at government involvement.
> The fantastic 7 tech companies, get billions of dollars from the government grants, tax breaks, business
> We have 4 ISP's that monopolize the market, due to the government giving them billions of dollars to expand infrastructure, which they misappropriated and then uses for lobbying against small ISP's
> The SNAP program subsidizes big companies like Walmart, target, etc due to them being able to eat the costs of all the massive regulatiom around it.
> etc. etc. etc.
That's not what capitalism is... or it's a very elementary understanding of it. Trade without force existed prior to and contemporary to capitalism. (See markets and mercantilism for more history there.) Capitalism is specifically the sort of mercantilism under which maximizing one's capital (typically the money commodity) becomes the aim.
Historically, money commodity was used to mediate exchange between two other commodities - I sell my wheat, get money, use money to buy pork. I'm buying commodities because I intend to use them or share them. Commodity -> money -> commodity. Capitalism inverts this -- the goal is to maximize money. I start with money, I convert it to a commodity that I can sell for a greater value, to get larger money. Money -> commodity -> money. You see Amazon operate this way: "if we can put one dollar in and get $1.01 out, do it".
Thank you for that correction, I did oversimplify it.
> Capitalism inverts this -- the goal is to maximize money. I start with money, I convert it to a commodity that I can sell for a greater value, to get larger money. Money -> commodity -> money. You see Amazon operate this way: "if we can put one dollar in and get $1.01 out, do it".
Yes, that is correct and there is nothing wrong with this as long as there is no coercion happening (eg. Government involvement, lobbying and over regulation).
Every person on earth should have the right to put in $1 into something and try to get $1.01 from it. You and I and every other individual get to then decide if we want to give that business owner that $1.01 or go to a competitor. In a truly free market, that business owner only earns that $1.01 profit by providing a good or service that voluntary customers value at more than the price they pay. The profit is the reward for creating value for others
We currently live in a crony-capitalist world which has a lot of issues, but it's still much better than the 200m+ that have died under Marxism and Communism in the last 100 years.
The whole article documents a product that people do want and like so much that they're willing to donate 4x what the owner-operators of the company asked for.
It's odd to see you expressing disdain for 'high school politics' and 'because younger generations don't know' while also displaying such ignorance or disregard of the subject matter, like watching factual waves roll in and dash themselves on the rocks of obdurate opinion.
The article is about a failing company on the verge of bankruptcy and which already went bankrupt 4 times this century, with turnover down 20% over the last 10 years...
My reply was to a specific comment. I do express disdain for those who preach socialism in 2025 because, as an European who was alive during the 20th century and has some mileage I do know the subject matter... And everyone should, too because of the ample historical and economic record available for all to study. Hence, "youth" is the only generous 'excuse' as the debate was conclusively settled by the 1980s. Judging by the puerile reactions I will go with that...
Please refrain from the typical ad hominem attacks and insults.
You realize there's an entire spectrum of socialist ideology between liberalism and authoritarian planned economies, yeah? It's not a Boolean, and it's not a slippery slope. We could have broadly liberal markets owned and run by the workers, and without the authoritarianism.
This is a slippery slope... The comment I replied to is an example of that: specifically on this aspect, it doesn't really matter whether a company is owned and run by the workers (we can have thay today and we do), but if it operates in a free market then it has to compete. Wishing that there wasn't competition is the slippery slope that ends badly, always.
Obviously mandating workers-ownership is authoritarian in itself.
Socialism can indeed only ends one way and that is the removal of individual freedoms.
It's not about eliminating competition. It's about trying to create an economic environment where companies are incentivized to compete by actually producing value rather than:
- Regulatory capture
- Buying out their competitors
- Being cheaper at the cost of horrible externalities
And the many other failure modes of "free markets".
Wishing that there wasn't competition is the slippery slope
Nobody is doing that. The article specifically mentions pursuing opportunities in the export market.
Obviously mandating workers-ownership is authoritarian in itself.
The workers bought out the previous owners, presumably due to frustration with the firm having had 4 near-collapses in the last 20 years under the capitalistic management approach.Your objections seem conditioned rather than considered; did you actually think any of this through, or are just just reacting to concepts that make you uncomfortable while overlooking the facts?
> but if it operates in a free market then it has to compete
There is no such thing as an actual "free market", nor should there be. Markets need rules to operate otherwise you get fraud and "unhealthy" competition like sabotage and assassinations.
> Socialism can indeed only ends one way and that is the removal of individual freedoms.
That's like saying that capitalism only ends one way and that is slavery.
In reality, there is no inherently slippery slope in either direction. There is a spectrum of ways to aporoach market regulation and individual freedom with various tradeoffs.
Blind partisan rhetoric like yours doesn't help uncover what those tradeoffs are so we can make good decisions. Instead it spreads ignorance and division.
Are you honestly comparing their proposal to the USSR and the "proletarian dictatorship imposed at gun point" to democratically changing laws to force all companies to become worker owned cooperatives?
At the end of the day, if we had a million worker owned cooperatives, they would still need to compete. Since workers from Coop1 need to sell better products than workers from Coop2, otherwise they will still lose their jobs.
Also, the current setup is a race to the bottom: lowest quality that doesn't self destruct immediately (aka planned obsolescence), dumping ALL externalities onto the general public: ideally with the same timed fuse approach so that it's not noticed immediately (dumping toxic waste in rivers, polluting the earth and the air, oil spills, etc - the list is endless), all in the name of a slightly lower initial price to attract customers. It's not sustainable.
The whole issue with this industry is that it moves so fast, there is no "long term." You're either in all the way in a likely futile attempt to capture the market or you're not in at all. So you also don't have time to really innovate on the hardware or software level and you need to put everything into training data and training hardware.
This. There's a fundamental logic error here. You simply don't hear about downtimes at smaller providers that often because it doesn't affect a significant portion of the internet like it does e.g. for AWS. But that doesn't mean they are more stable in general.
yeah, I'd like to see hard data on uptimes / reliability between these 2 services before declaring that big = bad and small = good.
FlyIO (and Digital Ocean) had horrible up-time when they first got started. In the last 6-12 months, FlyIO been much better. But they would go down all the time or have unexpected CI bugs/changes.
Digital Ocean accidentally hard deleted user's object stores before their IPO.
It's not like they're making their money from this though. All AI work is heavily subsidised, for Alphabet it just happens that the funding comes from within the megacorp. If MS had fully absorbed OpenAI back when their board nearly sunk the boat, they'd be in the exact same situation today.
They're not making money, but they're in a much better situation than Microsoft/OpenAI because of TPUs. TPUs are much cheaper than Nvidia cards both to purchase and to operate, so Google's AI efforts aren't running at as much of a loss as everyone else. That's why they can do things like offer Gemini 3 Pro for free.
A lot of major providers offer their cutting edge model for free in some form these days, that's merely a market penetration strategy. At the end of the day (if you look at the cloud prices), TPUs are only about 30% cheaper. But NVidia produces orders of magnitude more cards. So Google will certainly need more time to train and globally deploy inference for their frontier models. For example, I doubt they could do with TPUs what xAI did with Nvidia cards.
It actually works the same as on google. As in, ChatGPT will happily give you a link to a site with the lyrics without issue (regardless whether the third party site provider has any rights or not). But in the search/chat itself, you can only see snippets or small sections, not the entire text.
1. chatgpt is the publisher, Google is a search engine, links to publishers.
2. LLMs typically don't produce content verbatim. Some LLMs do provide references but it remains a pasta of sentences worded differently.
You are asking for gpt to publish verbatim content which may be copyrighted, it would be deemed infringement since non verbatim is already crossing the line.
Noone said it couldn't do that. In fact ChatGPT can do both. They just limit the direct content recital, because it is a weird area for copyright and Google also got burned for this already in some countries.
Same way the government needs to read all your emails because some terrorist on the other side of the world may or may not be using email as well to communicate.
Except that this other package also only came out last year and has contributed zero to Nvidia's current status. If AMD ever wants to be taken seriously in this market, they will need to start making their own software good instead of relying on "open source" in the mistaken belief that someone else will fix their bad code for free. Nvidia spent more than a decade hiring top talent and getting their proprietary software environment right before they really took off. And some of the older ML researchers here will certainly remember it wasn't pain-free either. But they didn't just turn the ship around, they turned it into a nuclear aircraft carrier that dominates the entire world.
Yeah honestly I’m dumbfounded why all these years AMD still doesn’t have an internal “code red” and get their developer experience up to par with CUDA.
Man if that is anything close to the truth, it would explain a lot and be pretty depressing. It would imply leadership doesn’t understand software at all, and considers it a liability rather than an asset.
It seems that AMD, like many other companies, doesn’t “get” software. It’s a cost-center, a nuisance, not really hard engineering, the community will take care of that, etc. It’s pretty ironic.
I found their acquisition of Xilinx (the FPGA company) to predict that they were going all in on a uniform FPGA / GPU / AI ecosystem, but… that didn’t seem to have yielded any integration benefits?
I’m genuinely dumbfounded by what’s up at AMD at this point.
reply