murderfs

> But I do care about, for example, a recently published author, or someone who posts their paintings on social media. Regardless of where the law stands in each particular case, I think it's repugnant that statistical models were trained on these inputs with the intent of offering the ability to produce similar outputs without any payment or even attribution to the original artist.

I have somewhat more sympathy for individuals, but I disagree with the entire system itself.

Suppose an aspiring artist learns from someone's work, and I commission them to produce similar output to the original without attribution. Was the act of them learning from the work "repugnant"?

> If someone chooses to make their information free, then so it may be, but the issue at hand is very obviously not about the use of, e.g., copyleft materials. The AI companies knowingly chose to take material that they knew they should not have and used it to make their copycat machines, building those machines in such a way that those responsible can cower behind the letter of the law to defile its spirit.

I don't agree that they shouldn't have done it. I disagree with the letter and the spirit of the law.

> 1. All modern LLMs are (in my eyes) inherently morally problematic. They were trained on stolen information, and where they didn't break the law they corrupted social contracts.

It's bizarre to me that over what feels like less than a year, enforcement of intellectual property and copyright laws has become the most important moral principle. Suddenly, it's your moral duty to enforce Disney's trademark of Mickey Mouse.

I reject this moral framing completely. Information wants to be free!

Yeah, my first reaction to your competitor analysis pricing is disbelief. I think it'd be a lot more convincing if you published your source files so that people could replicate your quotes.

It seems rather unlikely that solving the Navier-Stokes problem is harder than e.g. a better attack on a reduced-round AES. Perhaps they think an epsilon improvement in the exponent isn't worth publishing, but it seems more likely that they're intentionally withholding any results they have.

People keep repeating this, but it is only remotely close to true for the literal free plans and the small subset of consumers maxing out the pro plans.

At API pricing (i.e. what all enterprise customers are paying), inference is profitable, even if they're paying usurious rates on cost of capital for capex. Where the frontier labs are getting screwed is that they're in a Red Queen's race with each other, spending tens of billions of dollars per year training the next frontier model, and not being able to recoup the investment on that model before it's obsolete.

That's why you see the CEOs calling for government intervention to "pace the frontier": it lets them reduce investment on training without risking that their competitors keep going and make them obsolete. Unfortunately for them, the Chinese labs exist, and would love nothing more than to take the frontier, so they're in a bit of a pickle.

Pretty sure that's the eMudhra "private key is 4" issue.

Elliptic curves can be broken with 2^(n/2) steps, so you could probably do it on a single GPU within a day.

Vitamin B12 deficiency in vegans is a pretty obvious one.

But this is ignoring my point: if you go to the billions of people in the developing world and tell them they're going to have to switch their diets to soylent green because we, the first world, emitted too much CO2, they will rightfully tell you to fuck off.

> if you're honest to yourself you'll notice that plant-based nutrition is the only sustainable way to feed billions of humans on this planet

If you're honest to yourself, you'll notice that plant-based nutrition is not a sustainable way to feed billions of humans on this planet, because those billions of people would rather destroy the world than have some vegans enforce their morality on them while enjoying a lifestyle that consumes orders of magnitude more energy than some third world subsistence farmer's chickens.

It's not legal because it's unsafe for pedestrians, not for people inside the car.

To be clear, I don't believe either of these were wrongs. In my opinion, this is just a symptom of Facebook's PSC process training people to clout chase and demand recognition.

> Mainstream medical science along with common sense point to nicotine being addictive and bad for you.

Gwern's article cites many, many peer reviewed journal articles (aka mainstream medical science) about how nicotine's addictiveness is overblown, and you're citing three tweets.

Common sense points to you being clueless in all of your writing, even though I haven't read it, and I don't really feel the need to interrogate that view.