raof.cooperteam.net
He|They
337 posts
28 followers
40 following
Active Commenter
comment in response to
post
This is what I mean? It's “incredibly useful”, but also you can only use it for low-stakes ancillary stuff?
Maybe my experience is atypical, but that just doesn't describe the vast majority of my (or my friends') work?
comment in response to
post
It's not clear to me that it *is* incredibly useful? Even the boosters, like this piece, say “yeah, sure, it produces crappy code; I just have to review it”.
So, you've replaced the *easy* bit - typing out the code - by making the hard bit, that people suck at - reviewing the code - harder? WTF?
comment in response to
post
One of the really interesting technological features of LLMs is that they contain essentially a vector database of language, such that you can get the vector for “king”, subtract “man”, add the vector for “women”, and you end up near “queen”. This is interesting for conceptual exploration (search)!
comment in response to
post
The thing is, search is one of the domains *best* suited to LLM technology!
The trouble is that Google doesn't use them for *search*, they use them for *summarising the search results* which is a wildly different usage and something LLMs are inherently much worse at!
comment in response to
post
I mean, that has been their pretty explicit policy wrt Palestine for the last several decades.
comment in response to
post
Hey now, that's unfair.
They also want the state to ensure that their adult children come to Thanksgiving dinner, and don't make a scene.
comment in response to
post
Right!
Which is why Peelian policing emphasises that “that the police are the public and that the public are the police”.
“The Thin Blue Line”, police describing non-police as “civilians”, and so on are cultural practices of policing that must be ruthlessly stamped out to avoid ACAB.
comment in response to
post
ACAB, not because all cops are always bastards, but ACAB because if you don't deliberately, continuously, and with a lot of effort try to *prevent* cops from becoming bastards that is the inevitable state of policing.
comment in response to
post
Even in the best run system, police spend their time interacting with people at their worst. This is an environment that drives people towards bastardry.
comment in response to
post
This is true even for systems of policing that did *not* evolve out of fugitive slave patrols.
The London Peelian policing model was very deliberately constructed to avoid ACAB, but ACAB now describes the Met, too.
Because the job of policing inherently pushes towards bastardry.
comment in response to
post
It is possible to retain your idealism, to remain a non-bastard cop, but you start as a virtually powerless cog in a system of bastardry, and to remain a non-bastard you will frequently need to be in direct opposition to your colleagues and superiors.
It's possible, but lottery winners exist, too.
comment in response to
post
The problem is not that individual cops are bastards, or that only bastards join police departments.
The problem is that police departments are a machine for *making* bastards.
You join the dept of systematic problems as a rookie; you have *no* institutional power there.
comment in response to
post
The amount of good tooling work you could get done with $5b is *enormous*! OpenAI is a hideous misallocation of capital.
comment in response to
post
This is the other thing that I find frustrating about the LLM craze.
Your tools *could* be fixed, and you'd get a durable productivity bump. Good error messages from compilers are pretty common, nowadays.
Or! You could throw an LLM at it and get maybe a benefit, maybe a significant CVE increase.
comment in response to
post
You are worried about a phantom. *Inflation* is not correlated to debt levels, let alone hyperinflation!
There are plenty of economic problems to worry about. Hyperinflation is not one of them, and anything you do in fear of hyperinflation will make everything else worse.
comment in response to
post
The wildest variant of this is that the UK spends less per-capita (at PPP) providing universal free-at-point-of-service healthcare than the US federal government spends on Medicare + VA!
comment in response to
post
“Fixing” the debt will not improve anything you care about.
Fixing everything you care about will not (necessarily) effect the debt.
The debt is irrelevant.
comment in response to
post
I mean, yes, there are ways the Japanese economy is better structured than the US one.
But that's my point! The debt is irrelevant! It is not economically important!
comment in response to
post
The US federal debt does not affect¹ any economic indicator that you care about.
¹: Ok, not *quite* true. It affects *politics*, and the political response to the debt has economic effects. It has no direct *economic* effect.
comment in response to
post
This is just straight up wrong.
Government debt is not necessarily connected to inflation - as can be trivially demonstrated with Japan, which has had more than twice the debt/GDP ratio of the US for like 3 decades and has had lower inflation than the US over that time.
comment in response to
post
Fundamentally, the size of the debt is not economically important. It's not a constraint on what the government can do, it's not even meaningfully correlated with economic indicators that you care about.
In politics it is purely a rhetorical device, used to advance regressive policies.
comment in response to
post
Why would the US *want* to pay down the debt? *It would be a bad idea*. This is one of the fundamental differences between the issuer and users of a currency.
Also, the effect of inflation is not so straightforward. But, broadly speaking, inflation helps debtors and harms creditors.
comment in response to
post
But *is* the debt a problem?
“What, if any, relationship is there between a country's economic success and debt/GDP ratio” is an open question in economics.
comment in response to
post
The problem is that the analogy is false; governments that issue currency are fundamentally different to *users* of currency.
Also, the size of governments actually changes things; economics is not scale-invariant. You cannot run a country like a household, their economic constraints are different.
comment in response to
post
So, Sky News After Dark is highly influential on conservative politicians while only being watched by low-5 digit audiences¹. This is not a good way to be responsive to the needs of the general public.
¹: Their YT channel has quite an *international* following, but not domestic.
comment in response to
post
I think a significant factor in the AU result is compulsory voting. The conservatives exist in a media bubble, like in the rest of the Anglozone, but most people *aren't* in that bubble, and everyone has to vote¹.
¹: Or, at least, turn up to a polling place.
comment in response to
post
But it's important to remember that this is an aggregation; for example, the UK has a right to exist only so long as the citizens of the UK want it to. If (some fuzzily-determined supermajority) of Scottish UK citizens don't want the UK to exist, then the UK *doesn't* have a right to exist.
comment in response to
post
I think “Obviously (nation-state) has a right to exist” is a normally-helpful abstraction of “people have a right to self-determination”. *Most* of the time this comes up, people identify as members of (nation-state), so it's reasonable to aggregate those into “(nation-state) has a right to exist”.
comment in response to
post
I think this is incorrect. Does North Korea have a right to exist? Does *the UK* have a right to exist? Does the UK have a right to exist even if Scotland votes to leave?
“Does $STATE have a right to exist” is a straightforward “no”; states don't have rights, people have rights.
comment in response to
post
I'm not saying that's the *only* way to get consciousness, just that it's a guaranteed way to get artificial consciousness.
comment in response to
post
Like, cigarette advertising was banned in Australia *way* back, and now there's a lot less smokers around. (Although banning smoking in certain locations certainly helped there!)
comment in response to
post
Pretty easy to ban advertising of it, and that's a decent social signal that it's not really acceptable.
comment in response to
post
I agree that synthetic consciousness is theoretically possible - if by no other way, then by a sufficiently correct simulation of a human - but it's important to remember that terms used in AI development are *jargon*. Their meanings do not match lay English.
comment in response to
post
The problem is that neutral nets are analogous to human neurology in the same way that the motor of a car is analogous to a heart.
They are an architecture *inspired by* a dramatically oversimplified model of a *component* of neurobiology.
comment in response to
post
This is also sad because it'll diminish the (I think pretty good!) quality of the .NET code.
(Good) Code review is harder than writing the code in the first place! They've chosen to poorly automate the easy bit in a way that makes the harder bit harder!
All-round sadness for the engineers.
comment in response to
post
It's worse than that. The AI isn't a sadistic genie looking to grant your wish in the worst possible interpretation! There is *no* way to phrase your question to ensure the answer isn't horribly misleading, because it's fundamentally generating *text*, not answers.
comment in response to
post
I think the media environment would be significantly different if a balls-out lie like “they're eating the cats, they're eating the dogs” had a reasonable chance of resulting in a successful defamation suit. It doesn't prevent it; it *does* provide a chilling effect (in the direction you want).
comment in response to
post
And you could do much the same thing with hate speech - you could expand libel/defamation so that something like “Jews drink the blood of Christian babies” would be actionable group-defamation (*ideally* there'd be a public prosecutor of this, but you could also just do basically class-action stuff)
comment in response to
post
You don't have to prove that; you have to prove that they knew or should have known.
And I'm not saying it's *easy* to prove defamation, just that “the government should prevent *some* speech” is an uncontroversial opinion in the US. The difference with the rest of the world is *what* speech.
comment in response to
post
And also because the general public are *tremendously* misinformed about a whole host of issues (eg: polls show people think ~25% of government expenditures are foreign aid, and it *should* be ~10%. The actual spending is ~1%), in a way that not-coincidentally lines up with right-wing narratives.
comment in response to
post
I think it's also important to point out that “The government should prevent the media from nakedly lying” *is* within the US conception of free speech, for some types of lies. Notably, defamation and libel against people, a whole host of medical regulations, etc.
comment in response to
post
(Among other things, like “It is acceptable to ban political advertising at certain times”, and “it is acceptable to limit the amount of money spent on political advertising”, and things like that)
comment in response to
post
Because there's an entire news ecosystem dedicated to lying to their audience for partisan gain.
I think one of the reasons this is less the case in the rest of the Anglozone is that “the government should prevent the media from nakedly lying” is within the bounds of free speech outside the US.
comment in response to
post
This is significantly because there is a whole explicitly partisan media apparatus that freely lies to the public.
The Republican party can win despite being openly antidemocratic not because ‘I want to end liberal democracy’ is a popular policy, but because voters can pretend that isn't the policy
comment in response to
post
Not hate speech in particular, but more broadly a different conception of free speech (under which hate speech laws are considered appropriate).
Partcularly: the US information environment is polluted in a way that is exceptional (in degree, not type) in the Anglozone.
comment in response to
post
I think it's reasonable to wonder if the causality goes the other way: that one of the reasons their governments work better is *because* they have a slightly different conception of free speech.
comment in response to
post
At least in Australia it's not *super* common, but it *does* make “interest rates” a vastly more salient feature of the economic environment. A significant part of post-COVID inflation here was mortgage payments increasing as interest rates rose (to combat inflation 🤦♀️).
comment in response to
post
And, I mean, the Magna Carta was a bunch of stuff of the form “The king can't arbitrarily fuck with *his nobles* in this way”. As a step from absolute despotism, important. As a step to liberalism, it's like a baby finally being able to support the weight of their own head and look around.
comment in response to
post
I don't *think* Dutton is in the sort of electorate where “The type of Queensland cop that makes ‘Police Royal Commission’ float to the top of one's mind” is a serious electoral negative, but we can live in hope!