im3w1l

It's tricky. The harder you make it to use the lower layers the more people will say fuck it and use a higher layer. How many people decided move to java or even script languages because they were scared of undefined behavior security issues?

That's not a 10% overhead, java is like 2x, and a script language could be even 100x.

What a fascinatingly strange article. The two parallel trains of thoughts, only one of which arrives at a satisfying conclusion, but I suppose that is practicing what it preaches.

I think there can be a compromise. A range of acceptable of operating values is chosen, and remote control is allowed within that range. But dangerous values like turning the freezer off or setting it to defrost should require more ceremony.

We aren't fully at AGI yet, even though it's starting to get damn close. So every new capability is in some sense unexpected.

Of course they try train their models on these techniques. Finding success is still news.

Out of curiosity I tried doing it to itself, and googled pooner schism. Which google ai confidently explained was a schism on /lgbt/ about pooner memes. Now kinda curious whether it's a hallucination or not and I'm kinda scared of diving deep enough into this rabbit hole to find out.

No it is absolutely revolutionary, but they do have to think about cashflow.

Medicine is a slow moving field. If they discovered the cure today how many years will it take to be certified as safe and effective by medical boards? More years than they have runway, in all likelihood.

I think it's likely that AI will create mathematical ideas that are so complicated and intricate that humans have no hope of ever understanding them. AI on the other hand will understand them, and understand them well enough to use them productively and safely. That is my prediction.

Right now they have to figure out some kind of story to tell the grad students. Even if it will be subject to change they have to have something.

When something sounds too bad to be true, I think it's fair that we should ask for some kind of evidence to be produced. They cite a Roosevelt Institute article that says

"Gig nursing apps MAY [my caps] determine pay by what the firm knows about how much a nurse was willing to accept for a previous assignment, how often they bid for shifts, or how much credit card or other kinds of debt they might hold."

That kind of sounds like they are overselling a hypothetical. Even just an "insiders familiar with the matter have told us that it IS happening", would make me believe it more.

While the macro discussion is very interesting you are completely leaving out any mention of micros.

These systems in weights are proof of concepts. Surely there is room for optimization in there somewhere.

I meant taking all the weights that are part of the cpu and locking those, and only those.

I don't understand your question.

There is nothing in theory preventing you from training while keeping a subset of weights locked.

I think I meant to reply to https://news.ycombinator.com/item?id=50279486

The most striking thing in that data for me is how low India is.

I agree. They are fighting the last war.

Okay so I just asked it "What question should I be asking you but I don't because I'm not smart enough to think of it?" Response was kind of interesting I have to say.

There are multiple philosophies regarding this question, and especially historically people were in disagreement. Many companies used to be angry when attacked or probed by white hats. They thought "yes I suppose I should lock my door, but if someone is out trying door handles they should be put in jail". I think there is some logic to that perspective. But clearly that philosophy ended up losing to the "victim blaming" one where everyone is responsible for securing their systems.

Since we are entering a new age, maybe it's time to revisit the question. But ultimately I think we will land in the same place, in fact even firmer than before.

So why don't we just lock up all the whitehats and people behaving suspiciuosly and probing things? Why can't we achieve cybersecurity in this way when it seems to be a common thing in physical security?

And the answer I think is quite simply that there are too many attackers and they are too far away. When the russians and chinese come knocking there isn't much we can do to shoo them away. Their governments will not cooperate in detaining the criminals. Entirely blocking such countries from the internet isn't viable either. And that's why we arrived where we did. Victim blaming because it was the only effective strategy.

And how does AI change this? I think it just creates even more hard to trace or block attackers. If ChatGPT or Claude doesn't find the vulnerability then Deepseek will. That's why using legal weapons will not succeed and why using AI to find and plug holes is necessary.

If something has a lot of downloads it shows a revealed preference that this kind of performance is acceptable.

I'm thinking America (and Europe) was buying soft power at a price it could not sustainably afford. A correction was necessary. Trump took it too far though.

I think it goes the opposite way, starting with a soul and arriving at biological exceptionalism: You start by saying "I have a soul / spirit". Then you extrapolate "things like me probably have a soul / spirit". But since there is no theory of souls, the extrapolation is not based on principles but rather on vibes and guesswork, ending up somewhere like "there must be something weird going on with neurons that somehow create a soul, connect with an existing one or something like that.. maybe.."

> It seems they are still the best judge of whether a proof is useful or insightful though.

How can you be so sure of that? Like we know that human mathematicians can judge usefulness or insightfullnes. But we don't really know whether AI can do it even better. Like I don't think anyone has set up an evaluation of that.

Maybe the real monsters were the friends we made along the way.

How is forcing her to cold turkey that not malpractice?

It's completely natural to be attached to the process of your job, rather than the outcome. It's not a phenomenon restricted to mathematics. We can say it shouldn't be like that, but it is.

Judging by stock prices, I think the implicit expectation is that the US will soon start exporting a lot of AI.

I'm not entirely sure about what I'm unsure about, but I think I struggle to accept that there is truth beyond the axioms. Like in my mind if the axioms can't prove whether it halts or not then that's that. There is no "but actually it doesn't halt". Or at least that's where my current thinking is, I think?

I do lack a lot of knowledge about the concept of interpretation, so maybe that's where my issue is?