bigmadshoe

I don’t think it’s morally consistent to “warn” about the consequences while devoting your life to bringing about those consequences as quickly as possible. If I was in that position of power and truly believed what I was saying, I would devote my work to slowing down that process to give time for society to adapt, not speeding it up.

It’s more reminiscent of a religious group who smugly tells you that the end-times are coming, and only they are going to be saved. Except in this case they are literally bringing about the end-times.

It’s the same thing as always: with the wind of years of unlimited VC money in their sails, people at major AI organizations genuinely believe they’re smarter than everyone else. “Why do we need to do things ‘by the book’ if we’re so smart?”. “Move fast and break things” - except the thing they’re breaking is society.

We saw this with the non-stop flagrant messaging about how “AI is going to kill X% of all jobs”, as if saying the quiet part out loud wouldn’t have consequences worth considering. These people believe they’re omnipotent and thus untouchable.

One of the parent comments specifically mentioned “space engines”, not MMOs.

The improved token efficiency would only account for the output and reasoning tokens occupying context, right? What about the user messages and other context like file contents? I’m not sure how these measure up in a long conversation

My bad, apparently it's prescription-only for children in particular.

It's worth keeping in mind that melatonin is also prescription-only in many countries e.g. UK, Canada, Japan, Australia.

The problem with that take is that the evidence for melatonin is quite poor outside of jet-lag and certain more serious sleep disorders, and there can be unexpected effects elsewhere in the body when supplementing hormones, e.g. increased rates of depression for melatonin in particular.

Note the “but free from unnecessary inductive biases” part of my comment. By that I meant a decision to make each expert good at a human-defined thing.

Good point! I thought you meant splitting them and then doing inference with some kind of learned router while keeping all the split models loaded at once. What you're suggesting is pretty sensible.

Bitter lesson is knocking. Mixture of experts is essentially what you’re describing but free from unnecessary inductive biases.

Interesting. I guess I don’t see corporations as people, and thus I don’t see this as a violation of free speech.

Conversely it isn't very cool seeing the person who grew up as a McLaren fan dying prematurely due to being influenced to start a deadly habit from a young age. And I say this as someone who enjoys smoking.

What is the cost-benefit analysis that led you to that conclusion? I ask as someone who enjoys smoking but struggles to see any good argument for the return of tobacco advertising.

We can decide to pass laws to protect people from addictive behaviors. Do you support bringing cigarette advertising back for example?

Just use cursor based pagination. How is this any different whether it’s infinite scroll or separate pages?

Are you confusing 20th century with 2000s?

Aren’t we just donating the PFAS to potentially sicker patients?

Kerbal Space Program at a minimum

Absolutely. These things are inherently pointless - what makes them cool is the human ingenuity required to achieve them. Remove that and it’s totally uninteresting.

30k-60k out of >3m births doesn’t seem like a massive problem to me. Maybe the energy spent fighting this could be better spent elsewhere.

Do you have much evidence that this is a real problem on the scale of the US population? I would imagine a tiny percentage of citizens were anchor children (i.e. people whose parents came to the US purely to give birth to them, then left again)

British Columbia is far from the equator and just removed DST (permanent summer time).

They aren’t able to control whether the important facts people want to be updated on are positive or negative. But I find it less addictive to read the news with minimal colors and no images.

For US-centric news, I really like the text only https://text.npr.org/1001

Two things: 1) we have abundant training data for humanoid embodiments (watch humans do things), and 2) the world is already designed for humans.

I agree, but since we're talking about imagine understanding with text output, clearly a CNN is unsuitable. My previous comment was overly reductive and CNNs can still be SoTA depending on your performance metrics. I spent the earlier part of my career training CNNs, and they are very pleasant to work with.

CNNs are not SoTA anymore when it comes to large models, and also are not used to provide interpretations of images as text, but rather to classify, do semantic segmentation, etc.

You put it perfectly. And all these AI math startups don't actually care about mathematics. They are just using it as a proxy for general reasoning, with the VC pitch being some kind of world domination after they crack these problems.

I think it’s a part of official Claude code.

So it seems to go in capitalism sadly. If we could maintain healthy competition and avoid collusion, maybe we would be allowed to vote with our wallets. But right now that seems like a distant fantasy.