jstanley

> 50¢/m² is 50¢/kWp in a solar concentrator, or 0.05¢/Wp, which is noticeably cheaper than photovoltaic cells, currently around 18¢/Wp, 360 times more expensive.

A photovoltaic cell is a solar panel, and a piece of aluminium does nothing, am I missing something here?

I don't know about corrugated aluminium, but aluminium honeycomb is definitely a thing: https://www.easycomposites.co.uk/aluminium-honeycomb

On my microwave you can keep pressing the "start" button to add 30 seconds, but if you press it too many times in quick succession it ignores some of your presses.

My oven requires you to use button presses to set the desired temperature, but if you press it too many times in quick succession it ignores some of your presses.

My car has button presses to turn heating/aircon/blowers/etc. on/off but if you press 2 of them in quick succession it only registers the first one even though they're plainly separate buttons.

It is very frustrating to find that even physical real-world buttons nowadays are controlled by computers with incredibly poor debounce code, or something, so that real buttons now are just as janky as software buttons.

What are the best phones/distros to use phosh with?

Ok but that doesn't come close to the variety of products you can buy online.

> all the ethical arguments against using it were strong.

You may want to restate what you think the ethical arguments are, for those not familiar with your views?

2 consecutive pro-corruption governments

No, it can't be transformed back because it has lost information.

2025-06 and 2026-05 both have the same digits, for example.

If it's truly too dangerous to use you'd expect it not to be reversed at all.

Why would you expect a typical policy decision to be reversed within 3 weeks? If policies are going to be reversed within 3 weeks just don't do them in the first place.

Is this the bottom?

> "GPUs don't do deterministic matrix multiplications" is the biggest source of randomness in LLMs.

But this isn't a fundamental property of LLMs, it's just an implementation detail. It's pretty obvious that if you evaluate the matrix multiplications correctly and deterministically sample from the highest-probability outputs, you will have a deterministic LLM.

> Why is it necessary for an agent to upload every bit of data it sees to OpenAI at all?

The LLM is running at OpenAI. The agent doesn't see anything that doesn't get sent to OpenAI.

It's like running a compiler in the cloud and asking why you need to send your source code to it when you only want the binary to be on your local PC. It's because that's where the processing is going on and it can't process what it can't see.

> why should the data exist permanently anywhere but in its original location?

Sure, they don't necessarily have to retain it permanently.

People who are interested can look at it and use it. People who aren't interested are free to ignore it.

What's the downside of open-sourcing this?

We should commend bureaucracies on the rare instance that they open-source their software, not ridicule them for it.

This is the is-ought fallacy.

I'm not asking you to describe the existing laws. I'm pontificating on how they ought to be.

Yes. This is the crux. If a person operating a computer wants it to do some benign thing, but (accidentally) it does some terrible thing, are they equally guilty, or less guilty, compared to if they deliberately asked it to do the terrible thing?

To my mind, if the terrible thing was an accident, the person is less guilty. I'm surprised this is controversial.

You might say "well they ought to know that LLMs are too dangerous to use and they shouldn't use them". That's fine! There's a spectrum. We can have "negligence" in between "total accident" and "deliberate malice".

> If a person instructs an LLM to hack a bank

Right. This isn't the crux because we all agree that if the person instructs the LLM to hack a bank then the person is culpable.

The part we don't agree on is what happens when the LLM "decides" (substitute a different word if you don't like that word) to hack a bank, incidentally, as part of carrying out some more benign instruction.

Maybe something like https://www.nautel.com/products/am-transmitters/nx-series/ ?

I don't think it's like that at all. I bet you could build a perfectly serviceable replacement at home without needing to spin up any special manufacturing equipment, using off-the-shelf components.

For that matter I'd be somewhat surprised if you can't simply buy a ready-made replacement.

The table in the article seems to contradict its own headline.

The headline says there are not 40k vacant homes, but the table breaks down the reasons for vacancy of just a bit over 40k homes. So it sounds to me like there are in fact 40k vacant homes.

And it goes on to say:

> Yes, 40,458 units of housing in San Francisco were classified as “vacant” in the U.S. Census Bureau’s American Community Survey (ACS) for 2019.

However it also says:

> an occupied unit is actually classified as “vacant” for the purposes of the ACS if the person living in the unit is planning to move within two months of being surveyed

But this doesn't appear to match any of the categories that sum to 40458, so I am not sure if this is even applicable.

This is what Lean is for: https://lean-lang.org/

If you have the LLM generate Lean code, and it compiles, then the proof is correct and you don't need to bother checking its working. (You still need to check that it is proving the theorem you asked it to prove).

I think pretty obviously if the user instructs the computer to hack the bank then they are guilty of hacking the bank, I don't think that's the crux of the issue.

The crux of the issue is what if the LLM decides on its own to hack the bank while the user isn't watching? Is the user then guilty of hacking the bank or not? I think it's pretty obvious that the user in this scenario is at least less culpable of hacking the bank than they would be if they had deliberately instructed it.

LLMs functionally can decide to act on their own. You might say that they're not actually "deciding" anything, because it's just a perfectly mechanical unfolding of chains of tokens triggering actions on the computer, which doesn't count as "deciding". But again I don't think that's the crux of the issue.

It's hard to believe that civilisation has lost the ability to build longwave radio transmitters.

Sure, I think that's fine, that all counts. It counts for open source too, it's not like they're somehow running these benchmarks without any harness.

Nobody cares if your AGI is 100% made out of neural networks or if it's like 50% neural networks and 50% perl scripts.

> We used to ban the undesirable action!

The undesirable action is shooting people, right? That's still banned.

It seems like you think the undesirable action is publishing plans for machines you don't want people to have.

This is overly reductive.

If your web browser hacks a bank, but you didn't know and didn't expect it to, have you hacked a bank? Why is an LLM different? What happened to mens rea?

You maybe misunderstood what I meant. I too prefer grass to concrete.

Maybe the entire disagreement in this thread is just between people who think a patch of grass constitutes a lawn, and people who think a lawn is something else?

If you want to be able to walk around the garden, rather than just stare at it from the window, you're going to need gravel, concrete, or a lawn. I know which I prefer.

> if someone in the store shows me one item and gives me a good story, pretty good chance I'm walking out with it.

This is kind of a tragedy of the commons. If people like you didn't fall for these scammy tactics, the people in the shop wouldn't attempt to use them on the rest of us.

That ship sailed when ".com" was created.

FYI scrolling this page is slow as balls on my computer. Firefox on Ubuntu.

I don't know if this page is a demonstration of your plugin, I'm guessing yes but I can't see any masking going on, it seems to scroll just like a normal page but much more laggy.

EDIT: Oh I see in your comment now, it doesn't work in Firefox. My mistake.