coldtea

>Just because the intelligence is artificial now doesn’t really change the answer.

Not only it changes the aswer, it's literally the whole of the question. When people talked of the posibility or imposibility of “natural language programming” they *always* meant "natural language instructions which *a machine* translates to running code".

They didn't have to spell it out, since nobody doubted a human programmer can write code when given natural language instructions - that was already a thing since ENIAC, if not since Ada Lovelace.

That's what we have now, and that's what changed between now and 1978 when considering whether “natural language programming” is a thing.

>We are what ~3 years into mass adoption of llm tech? Every couple of months it improves.

Nowhere as much as 1-1.5 year ago.

And "it's just 3 years" misses the fact that the spending for it (VC money, investment, etc) has been 10 to 100 to 1000 times what other technologies got for many years in their grown stages, just to get where we are.

They didn't chose to break user boundaries. They chose not to artificially recreate them.

In the shared daemon model, there are no user boundaries. Even if the files belonged to each user permissions-wise, the central daemon running as root would be able to edit all.

>Why doesn't docker respect unix filesystem permissions? I use docker on a shared server, i can delete docker images of other users. This is bonkers.

Same reason a Postgres root can delete tables of other users.

It could, but it would be a shim on top of Docker to respect them.

Docker is a service daemon running as root, and the images of other users are just entries created by the other users in its registry, not server files that have permissions of each user.

When you're creating a new docker image, you just send a command to that daemon process - you're not passing your user account credentials to create. As far as docker is concerned, it's just one more image it knows, not an image belonging to you specifically.

>If AI is natural language programming then they already had it in 1978, because AI is in fact "asking a programmer to write code for you"

That would just be the "I" part, since was missing the A for artificial: you asked a naturally made person.

And yes, nobody argued that they didn't have I in 1978, so I'm not sure what the argument is supposed to prove.

>Guess what wasn’t? The nonfunctional requirements.

OK, so you test it, see which nonfunctional requirements it doesn't meet, and ask the AI to change the code to include them - repeat until you're happy.

>Dijkstra's arguments tend not to change their minds.

Maybe they trust the practical results they see over what a professor wrote 50 years ago as an essay.

Conveniently enough in the future than most wont care if it doesn't pan out, but also soon enough to help push AI and its IPOs...

Humans stop learning once their training is finished too, it's called death.

The difference is that LLMs we can talk to them when they're in that dead state.

And of course for an LLM this is an implementation detail. One could just as well keep updating the weights of the same model over time. And its being done, it's called "continual learning". LoRA is also a way of doing that.

>Gettier's conception of knowledge as justified true belief (prob familiar to many philosophy undergraduates) implies that machines can't in fact accrue knowledge.

Nothing about the conception of knowledge as "justified true belief" implies that machines can't accrue knowledge.

For that to be the case "belief" would have to be defined in a circular way to exclude machines (which we can reject), or "justified" would have to be restricted to mean "based of observation of the external world".

The more general "based on observation of their input" applies to both men and LLMs.

That man can't know the bare direct facts of the world, just his own input (it's sensory experiences, and whatever second hand input he's given by other people), and for neither can he be sure they're absolutely true, or that he got right.

>The idea that knowledge can't be defined in a non-circular way is only plausible for someone with no familiarity with epistemology.

The idea that this or that epistemology restricts what can be the case in reality is only plausible for someone with no familiarity with epistemology.

Maybe those researchers are learned enough to understand that it's still interesting to know what a probabilistic database will pick among billions of words of training, millions of books and posts and videos and that stuff, and especially interesting that's there's so much consensus between them, while their approaches to many other questions are very different.

Or they might be learned enough to consider that our brains might also just be prediction machines, meaning probabilistic databases of experiences and facts we've learned, and the only thing we do is reproduce patterns more or less abstract in function of the pressure our experiences put on certain representations during our lifetime.

It's the "Author did shit all" prize

agent-scale is the new "Mongo is web-scale"

>Drum machines are instruments. Even if sequenced, drum machines are live instruments and they require skill.

Only if played live via it's pads or hooked to MIDI drum triggers.

If it's sequenced from some computer program like Logic or through step input, then it's not used as a "live instrument".

>I never play the same thing twice

99% of tracks however do. Are we discussing what rarely happens, or what's the most typical case in pop music, what TFA covers?

>Americans aren’t just moving, they’re bringing billions in wealth with them. (...) These inflows aren’t just large—they’re high-income.

AI slop everywhere.

>Did you read the quote you quoted?

Did you read what I wrote? Did I say it wasn't created with her SSN? I'm saying "if it was created with her SSN, then the service should have given her access.

We use the "if X" in my parts to mean "since X was the case, then....". And it seems it's valid in US English too:

Even if the speaker has just read a report explicitly stating that X did Y, using "if" is perfectly possible. The important distinction is that "if" does not necessarily indicate doubt about whether the Y occurred. In this context, it can have two interpretations:

Conditional justification: "If X did Y, then so and so is justified" The speaker is explaining what makes some reaction justifiable, without necessarily questioning the report.

Uncertainty about the facts: "Assuming X really did Y, then so and so was justified." Here, the speaker is reserving judgment about whether the report is accurate.

Both interpretations are available to a native English speaker.

>Even if she can (which is probably the case), she wants to go through discovery

The case, as per TFA, was the opposite: she wanted to get access and couldn't, and she finally managed through discovery.

You describe what she "should have preferred" (in order to have the trail), I was talking about the actual case.

>Smith later discovered why she had been arrested: a scammer had tricked someone into sending $1,000 to a Cash App account, one that was in her name but controlled by a criminal.

And that brings $50,000 bail and days spend in jail before trial? Should just be a civic matter at worst.

>It turned out that the account had been created using Smith's Social Security number and date of birth. But Smith, who describes herself as not tech-savvy, said she had never used Cash App. The situation sounds like one that should have been resolved quickly. But Smith says she is still being asked to prove her innocence in an ongoing criminal case. One stumbling block was that Cash App refused to hand over details of the account to Smith because she couldn't prove she was the owner. She eventually received the data through discovery in her case.

Surely some details about her are the same as in the account profile, and not just the name, else they wouldn't prosecute her for an app account that bears no relation to her. Do they prosecute people just for something happening on the same name and surname as them?

If the account had others details about here to fool people, like her photo or SSN or address, she should have got access to the account, without having to demand it via discovery.

>The entire premise of this essay rests on the idea that musical performances are necessarily played live (particularly, with a live drummer), in the tradition of rock music. This just qualitatively does not apply to much modern music, where drums are played by a machine or computer

One could say modern music is not "musical performances" then. A sequencer or drum maching playing isn't "a performance", it's a programmed loop.

>Do those genres 'groove' worse and feel less human?

Yes.

>Why would losing the ability to play together cause the kick and bass to line up?

Because if they lost the ability to play together, then they would be forced to either overdub tracks played individually with a click track, OR use midi, in both of which cases the kick and bass typically lines up.

So what this say, is we get this exact "perfect line up" now, but not because musicians lost the ability to play together and sound naturally, but because we prefer these other two production techniques and audio processing that removes all natural nuance.

>Odd comparison ... code was always malleable

Not really in the way the comparison means it. Code was malleable with big force (mandays of manual coding to change something). Now it's malleable with "change this to do that" on an LLM.

In the analogy, the former is more like a stone is also "malleable" (you can take a chisel to it after all), but much less than clay, and is a much more manual process.

>completely failed at its intended purpose: a way for non-programmers to ask questions of a database

And yet, non-programmers, like data analysts, had corporate roles doing exactly that.

>The compiler was supposed to end programming too

Bad analogy. The case with AI is not the same.

>then, which model does it, or if any does it now.. doesn't matter. what matters is that some inevitably will, and which models you pick will matter.

That's an abstract point, that might or might not be something to actually fear.

The article makes a concrete one. Point being it's both difficult to prove, and they have a motive to sell their solutions.

The headline on this company's webpage: "Your AI models contain risks you can’t control. Until now".

Their business model is selling AI danger and fixes for it.

I'd take their findings with a lot of salt.

Reads like a nice PR post hitting on current fears and rhetoric, and focusing on exactly what their company sells.

FUD and/or marketing: they're simply promoting their alignment/"westernization" services.

> Your AI models contain risks you can’t control. Until now.

Their business model is selling AI danger and fixes for it.

I'd take their findings with a lot of salt.

>This is an earnest question but is that really true for reactivity and state management?

Yes. We could do those before React and other crap too.

I'm showering you with negativity

OK, so it IS "taken away from them" as their job, which was the point.

Or did you think the argment was that they'll be forbidden by force to do math, and come to refute that?