MalikTerm

As you seem to be unaware of where the 12,183 arrests figure comes from, and suggest that you haven't seen compelling evidence for this figure, you should know that it's from The Times(0). They found that this led to 1,119 sentencings.

In your linked post [1] you suggest that this figure is completely wrong. To demonstrate this, you linked to a FOI request for the Metropolitan Police which shows that the actual figures for 2023 are 124 for Section 127 of the Communications Act and 1,585 for Section 1 of the Malicious Communications Act. This is, ironically, completely wrong. These figures are only applicable for the Metropolitan Police in the Greater London area, if you want figures for the UK you need to file FOI requests for all 45 territorial police forces in the UK. This is what The Times did, and 37 of them responded.

The Standard(1) attempts to address the claim of whether the UK arrests more people for social media posts by looking at figures from other countries, fails to point out a country with more arrests for social media posts, and concludes that open and liberal societies will have more arrests for social media posts because we are more free to do so. Go figure.

You also suggest that racial harassment, domestic abuse, stalking, and grooming are covered under the law, which is somewhat true, The Times quoted a spokesperson from Leicestershire police which stated that the laws cover any communications and may deal with cases of domestic abuse, and this is often the only example given to explain the figures. However it should be noted that the Communications Act(2) only covers electronic communications that are 'grossly offensive or of an indecent, obscene or menacing character' (or posts a message known to be false for the purposes of causing 'annoyance, inconvenience or needless anxiety', prior to 2024), and the Malicious Communications Act(3) while covering letters, EC, articles, etc, only applies if the communication is 'indecent or grossly offensive' (or a threat, prior to 2024).

For some of those issues it can be easy to point to communications that are 'grossly offensive' or threatening/menacing, however there are other more applicable laws to choose from such as the Public Order Act, the Crime and Disorder Act, and the Domestic Abuse Act which largely covers hate crimes and domestic abuse. An order of magnitude more people are arrested for hate crimes under these and similar laws than they are for malicious communications. The Protection from Harassment Act which covers harassment and stalking, the Serious Crime Act covers controlling and coercive behaviour, the Criminal Justice and Courts Act for revenge porn, and the Sexual Offences Act which covers an incredible amount of offences (it's a large act), including everything related to grooming.

The CPS largely discourage using communications offences (unsourced, but (4) is a good starting point), possibly because of the mens rea requirements for 'grossly offensive' or causing distress or anxiety, possibly because the sentencing limit for either communications act limited at 6 months or 12 months for malicious communications (also 6 months for offences prior to 2022), possibly because it has to weigh whether the sentencing is within the public interest with regards to the chilling effect it can have on speech, especially when concerns about Article 10 of the ECHR are brought up, but it has recommended using these acts as a fallback. Prior to 2015 revenge porn wasn't a specific offence but could still be considered under the communications acts for instance.

All of this to say, if the communications acts are being used as a fallback for the issues you mention, it can't be seen as anything other than a failure that the more specific legislation fails to address issues of or prosecuting issues of 'grossly offensive' or 'threatening' communications appropriately, which seems unlikely, but if it is the case, why then is the sentencing rate so pitiful? 10%? For 'racial harassment' and domestic abuse? In a country that records around 130k hate crimes and 230k cases of domestic abuse (of which 35% are related to malicious communications, do the maths) yearly? When the bar is as low as racial slurs or 'threats'?

For a number of high profile cases you could perhaps make the case that the arrest was justified, but these cases are high profile for a reason, they're testing the limits of what can be considered 'grossly offensive' that aren't covered by other more applicable laws. But even then, there are high profile cases simply because the police had absolutely no business arresting anybody(5). For it to be the case that these laws, specific to 'grossly offensive' and 'threatening' behaviour, are being used to address these issues it needs to be demonstrated, and I don't think that has been the case, and the issue of wasted police time needs to be addressed when 90% of arrests didn't need to be made. The last point is especially relevant at a time where petty crime has all but decriminalised over the past decade and when police chiefs are suggesting citizens are the ones that need to do something about shop lifters(6).

In the greater context of the conversation, it should be obvious that police are arresting people for social media posts, regardless of whether you agree with the intent or not, and it should be obvious that the police are interested in policing social media given the absurd number of Non-Crime Hate Incidents being recorded, also around 13,000(7) a year, and I can't see things getting better with the introduction of the OSA. Blaming these issues on a 'right-wing narrative' seems naive at best and missing the forest for the trees at worst. Labour having absolutely abysmal polling issues should suggest that this isn't a partisan issue in the slightest.

(0) https://www.thetimes.com/uk/crime/article/police-make-30-arr...

(1) https://www.standard.co.uk/news/tommy-robinson-uk-speech-cla...

(2) https://www.legislation.gov.uk/ukpga/2003/21/section/127

(3) https://www.legislation.gov.uk/ukpga/1988/27/section/1

(4) https://www.cps.gov.uk/prosecution-guidance/communications-o...

(5) https://www.bbc.com/news/articles/c4gz1qy30v5o

(6) https://www.lbc.co.uk/article/its-up-to-the-public-to-stand-...

(7) https://hansard.parliament.uk/lords/2024-11-19/debates/0DE7E...

I thought the countries you listed as maintaining book bans was a little short, given that basically every Western/First World/Global North country has previously banned or continues to ban books. For instance, Belgium, Finland, France, The Netherlands, Norway, Poland, South Korea, and the United States have all banned the sale and publication of books within recent memory, with restrictions on the importation of books and potential detention and fines for people importing them, and there may have been legal challenges made to the authors and publishers of books.

The definition you gave for banning books does narrow it down a bit though. Restriction on the publication, sale, or importation isn't a ban, but possession is. Ah. I guess a book is only banned if you face legal consequences for owning it after purchase during the very limited window that it may appear on bookshelves. In that case, I can't say for certain that Australia, Canada, Germany, and New Zealand bans books, but I can say that I would not want to be in possession of certain types of books in these countries. The definition of 'banned books' gets a little murky here though, is the inevitable police detention and interrogation due to the possession of the books themselves or rather obscenity/hate speech/Nazism/terrorism/drug laws? Does it count if the detention and interrogation occurs after an unrelated search of your property? Is a book only banned if it is confiscated from your possession? Does it count if you win your court case?

Which leaves the only Western nations that fit such a definition as being Austria and the United Kingdom, which do objectively criminalise the possession of banned books.

>It's because English has no (or very few - I can't think of any) words that begin with the same phoneme.

Loan words, but: Tsar (zar or sar), Tswana (50/50), and Tsetse fly (usually /ts/) from the Tswana language. I don't think /ts/ ever refers to something specific in native English, it's usually plurals like it-s or from suffixes like bet-sy, gats-by, wat-son.

It's an interesting choice to suggest that the switch to Hepburn romanisation was motivated for a desire to better help English speakers pronounce Japanese words when tsunami is your example. The official Kunrei-shiki romanisation for つなみ is 'tunami', and I can promise you that nobody who visits Japan tells their friends and family that they visited Mount Huzi (ふじ). You would have a point if you had chosen something like Mitutoyo, but even then names are usually the exception when it comes to romanisation/anglicisation as official rules are less applicable, cf. Mitsubishi.

Still, something like 'sooonami' is particularly grating even if we ignore the pretentious BBC accent (I have heard tsu-na-mi on BBC shows to be fair). It could be because as you said the onset gets simplified to better fit English phonotactics like with other words: (ph)thalic acid, (p)terodactyl, kr(w)asan (croissant) in American English with a doubly 'wrong' t at the end, (k)nife, (g)nome, sometimes (g)nu, etc, but I don't think this is it. Su-na-mi sounds fine and this is how it's pronounced in Spanish and some other languages too, every language ends up 'mispronouncing' words if it doesn't fit nicely into the existing phonology. I think what bothers me the most about 'sooonami' is the stress inevitably gets placed on the second syllable which becomes 'nah' in non-rhotic accents which just sounds wrong, and in terms of Japanese phonology it's rare to place the stress on the middle syllable, never mind that the mora is wrong and the pitch accent is wrong, but I by no means speak Japanese.

As for why English even uses tsunami in the first place, maybe 'tidal wave' makes sense if that's what you grew up with or you live in a part of the world at risk of tsunamis, but I don't think I made the connection until I was an adult. Are all tides not waves? Tidal bore, tidal flood, storm wave, etc, sure, unusual events relating to the tide or weather, tidal wave fits if we ignore that they're not caused by the tide, but it doesn't seem comparable to me even if tidal wave isn't wrong and is synonymous.

Q4 is usually around 4.5 bits per parameter but can be more as some layers are quantised to a higher precision, which would suggest 30 billion * 4.5 bit = 15.7GB, but the quant the GP is using is 17.3GB and 19.7GB for the article. Add around 20-50% overhead for various things and then some % for each 1k of tokens in the context and you're probably looking at no more than 32GB. If you're using something like llama.cpp which can offload some of the model to the GPU you'll still get decent performance even on a 16gb VRAM GPU.