Interesting Links for 04-06-2019

[andrewducker]

Autism: "What If Trains, But Too Much?"

(tags:children autism parenting behaviour )

Military rules, by service (funny)

(tags:military funny )

"10,000 Steps Every Day" is a health myth

(tags:health walking myths )

Once again, a neural net tries to name cats

(tags:cats names Tumblr ai )

I need to know what happened in this house!

(tags:housing Edinburgh wtf )

Who remembers cat macros?

(tags:cats meme )

Caster Semenya able to run medication-free for now as Swiss court floors IAAF

(tags:hormones gender sports Switzerland )

Helsinki's radical solution to homelessness

(tags:housing homelessness Finland )

France Bans Judge Analytics, 5 Years In Prison For Rule Breakers

(tags:france law analysis wtf censorship )

Project Details - Meadows to George Street pedestrianisation/segregated cycleways

(tags:edinburgh transport bicycles )

What are the Tory candidates' Brexit policies?

(tags:Conservatives uk europe )

Comments

[rmc28]

There is definitely some kind of story behind that house! (Catastrophic water tank failure?)

[andrewducker]

Pot growers?

[movingfinger]

No photos of the upper storey! I'm wondering why they left any of the plaster on the walls at all. If it's mold, they'd really need to. For standard water damage, maybe not? Those look like all new floorboards.

[azdak]

That house is straight out of Tana French's Broken Harbour.

[xenophanean]

Have you noticed that although the AI cat names are still weird and funny, they're steadily becoming better cat names.

Tom Glitter, Notion, Scarlet Be Thy Coat; these are all pretty good names for a cat.

[movingfinger]

I might not NAME a cat "Retchion," but I would CALL one that, for reasons.

[skington]

politics.co.uk isn't the only site to do this, but it's the one I noticed just now: if you click on the “do you want cookies?” link, and say “reject all” and save, it then redirects you to a “but cookies are nice!“ page. (Other sites redirect you to the home page instead.) This is pointless and petty, because you just need to click your back button to go to the article you wanted to read.

[skington]

I think on balance France has got it right on judge analytics, although more thought is needed. Presumably if companies can data-mine their way to working out individual judges' foibles, they can then sell that data to lawyers in front of those judges, influencing the lawyer's strategy, and at that point the principle of equality in front of the law has taken yet another pounding, as one side can afford to pay for information and the other side can't.

The alternative - to make that information useless rather than illegal - is to constrain judges' decisions so they all end up performing identically. And how would you do that? Presumably, by building exactly the same sort of machine-learning models that the analytics companies are building. But you have the problem that you're then using a very-potentially-flawed and difficult-to-understand computer system, not simply to inform a lawyer's decision-making process, but to make an actual legal decision. If the computer system is faulty, a lawyer can decide to ignore it in some cases and/or to not bother using it any more; if it's baked into standard court administrative workings, it's going to be much harder to unpick.